Structural basis for pharmacotherapeutic action of triple reuptake inhibitors

Abstract

Most first-line pharmacotherapeutic strategies for depression aim to boost serotonin and norepinephrine levels. However, 35% of patients with depression do not respond adequately to these treatments or experience adverse side effects. The serotonin–norepinephrine–dopamine reuptake inhibitors, also known as triple reuptake inhibitors (TRIs), are emerging as promising antidepressants with greater potency and fewer side effects. Here, we determine an ensemble of structures of DAT in complex with five distinct TRIs. Tesofensine and dasotraline stabilize DAT in an outward-facing conformation, while centanafadine, ansofaxine, and nefazodone capture the inward-facing conformation. These structures reveal binding poses and interactions involved in the association of inhibitors. Notably, ansofaxine binds at a location which is much closer to the intracellular membrane surface. Through extensive structural analysis, we establish a comprehensive blueprint for the association of these TRIs, which is crucial for future drug development aimed at achieving potent antidepressant with fewer side effect.

Most current antidepressants target SERT and NET but often have limited efficacy or side effects. Here, authors reveal how five triple reuptake inhibitors bind to DAT, offering structural insights to guide development of more effective antidepressants.

Introduction

Dopamine, a monoamine neurotransmitter, is released by dopaminergic neurons and acts on specific target cells to perform multiple physiological functions^1–4. Dopaminergic neurons are located primarily in the substantia nigra pars compacta (SNc), ventral tegmental area (VTA) and hypothalamus, sending projections mainly to the striatum and the nucleus accumbens (NAc) to regulate motor functions, appetite, motivation, emotion, reward association as well as addiction^5,6. Dopamine in the synaptic cleft undergoes reuptake into presynaptic neurons via the dopamine transporter (DAT), effectively regulating the intensity of dopamine-mediated signaling^7,8. Disruption of dopamine synthesis, metabolism and transport, especially reduced levels of dopamine, is associated with diverse neurological and psychiatric disorders including Parkinson’s disease, depression, schizophrenia, and attention deficit hyperactivity disorder (ADHD)^9,10. Conversely, interventions to rebalance dopamine homeostasis are widely used to treat related diseases and DAT serves as a common therapeutic target in this strategy¹¹.

Depression, a complex and heterogeneous mental disorder, affects hundreds of millions individuals worldwide, making a substantial contribution to the global disease burden^12,13. The monoamine hypothesis of depression suggests that the major depressive disorder is associated with low levels of monoamines in the synaptic cleft^14–17. The first generation of antidepressants, including tricyclic antidepressants (TCAs), named for their three-ring chemical structure, and monoamine oxidase inhibitors (MAOIs), is now largely restricted in use due to off-target pharmacology, poor tolerability, and susceptibility to drug-drug interactions^18–22. So far, selective serotonin reuptake inhibitors (SSRIs) and serotonin-noradrenaline reuptake inhibitors (SNRIs) are the first-line medications for treating depression, primarily aimed at increasing the extracellular level of serotonin or norepinephrine^23–25. However, these antidepressants do not provide therapeutic benefits for all patients, take approximately 2–4 weeks to achieve therapeutic effects, and may cause intolerable side effects, including anhedonia, weight gain, and sexual dysfunction^25–28. Accumulating evidence have provided compelling evidence that dopaminergic neurotransmission plays a pivotal role in depression^29–31. For instance, DAT-knockout mice display antidepressant-like behaviors³², and the concentrations of homovanillic acid, the major metabolite of dopamine, are significantly reduced in the cerebrospinal fluid of depressed patients^33,34. Moreover, the norepinephrine-dopamine reuptake inhibitor (NDRI) bupropion, when combined with SSRIs or SNRIs, not only enhances symptom improvement over monotherapy but also mitigates the side effects of sexual dysfunction commonly associated with SSRIs^35–37. In addition, the combination of SSRI citalopram and NDRI methylphenidate can significantly accelerate the onset of action in elderly depressed patients³⁸. Therefore, triple reuptake inhibitors, that simultaneously inhibit the reuptake of dopamine, norepinephrine, and serotonin, have emerged as promising and highly regarded antidepressant options.

In recent years, several TRIs such as dasotraline, tesofensine, centanafadine, and ansofaxine have been developed to treat depression and other disorders associated with monoamine deficiencies, including ADHD, and obesity^39–41. These are currently undergoing clinical trials and have shown improved therapeutic efficacy. For example, ansofaxine has shown a rapid and significant antidepressant effect, particularly effective in reducing anhedonia without raising risks of suicide, weight gain, or sexual dysfunction^42–44. Centanafadine also effectively manages core ADHD symptoms, with a favorable tolerability profile and a quick titration phase^45,46. However, the mechanisms by which TRIs bind and inhibit all three monoamine transporters, and how their binding modes differ from those of selective and conventional SSRIs and SNRIs, are poorly understood. This information is crucial for further drug design targeting monoamine transporters (MATs) and other closely related neurotransmitter transporters.

In this study, we obtain high-resolution structures of DAT bound to TRIs, including tesofensine, dasotraline, centanafadine, ansofaxine, and nefazodone. The structures, either in the outward-facing state or inward-facing state, reveal the multiple drug-binding modes of DAT, which were further validated by [³H]Dopamine transport assays. These structures enhance the understanding of the inhibition mechanisms of TRIs and provide insights into the development of antidepressants against MATs.

Results

Structural determination of DAT in complex with various TRIs

To investigate the binding mode of TRIs with human DAT (hDAT), the N-terminal 56-residue truncated hDAT, previously characterized in earlier studies^47–49, was expressed in HEK-293F cells and purified. Prior experiments have demonstrated that there is no significant difference between the truncation and the full-length DAT in terms of substrate transport and ligand binding⁴⁷. The construct was henceforth referred to as DAT unless otherwise specified. Subsequently, we purified the DAT protein using Lauryl maltose neopentyl glycol (LMNG) plus cholesteryl hemisuccinate (CHS). The purified protein was then reconstituted in nanodiscs consisting of brain polar lipids (BPL) and cholesterol (Supplementary Fig. 1). During the purification process, specific TRIs, such as tesofensine, dasotraline, centanafadine, ansofaxine, and nefazodone were introduced to obtain the complexes. For each complex, we then froze the grids, collected cryo-EM data, and determined the structure at resolutions ranging from 2.8 to 3.5 Å using cryoSPARC⁵⁰ (Fig. 1, Supplementary Figs. 2–6 and Supplementary Table 1). Similar to other neurotransmitter sodium symporter (NSS) members, five complexes adopt a classical LeuT-fold with 12 transmembrane helices in which TM1-5 are related to TM6-10 by a pseudo-twofold symmetry (Fig. 1b, c). Notably, TM1 and TM6 feature a non-helical conformation in the central region of the transmembrane area, where ligands and ions bind through numerous interactions. The state of the central region is also an important criterion to determine the conformation of the transporter. Among the structures we obtained, tesofensine and dasotraline stabilized DAT in an outward-facing conformation, while centanafadine, ansofaxine, and nefazodone stabilized DAT in an inward-facing conformation (Fig. 1b, c).

Fig. 1. Cryo-EM structures of human DAT bound to triple reuptake inhibitors.

a Densities and chemical structures of five inhibitors within the structures of DAT complexes. The cryo-EM densities of dasotraline, tesofensine, centanafadine, ansofaxine, and nefazodone are shown in the contour levels of 0.28, 0.32, 0.28, 0.25, and 0.2, respectively, as visualized in ChimeraX. b Side-view representations of the cryo-EM maps for the DAT-inhibitor complexes. c Overall structural models of DAT in complex with five inhibitors, with the inhibitors depicted as spheres.

Binding mode of tesofensine

Tesofensine is a triple reuptake inhibitor currently undergoing phase II trials for the treatment of obesity^51–54. It exhibits inhibitory effects on DAT, NET, and SERT, with reported IC₅₀ values of 6.5 nM, 1.7 nM, and 11 nM, respectively⁴⁰. We determined the structure of DAT bound with tesofensine (DAT^teso) at a resolution of 2.8 Å (Supplementary Fig. 2 and Supplementary Table 1). Notably, a cloverleaf-shaped density was resolved, consistent with the molecular structure of tesofensine (Fig. 2a, b). Tesofensine consists of a tropane ring, a dichlorophenyl group, and an ethoxymethyl group, and is cradled by TM1, TM3, TM6, and TM8 helices, located 15 Å deep from the extracellular membrane (Fig. 2a). Its binding pocket is exposed to the extracellular side, indicating that DAT^teso is stabilized in an outward-facing conformation. The tropane ring is positioned between TM1 and TM6 and is surrounded by F76^TM1, D79^TM1, and F326^TM6, with the amino group of the tropane ring forming a salt bridge with the carboxyl group of D79^TM1 (Fig. 2b). The ethoxymethyl group points toward the extracellular side, in close proximity to F155^TM3 and Y156^TM3. The phenyl ring of the dichlorophenyl group engages in a T-shape π-π interaction with Y156^TM3 and is further enveloped by S149^TM3, V152^TM3, and S422^TM8 (Fig. 2b). The mutations F326A and S422A, although they retain some uptake activity, substantially decrease the inhibitory potency of tesofensine (Fig. 2c), consistent with our findings that these residues are crucial for the binding of tesofensine. In addition to these direct interactions, the short side chains of G323^TM6, G425^TM8, and G426^TM8 are critical in providing sufficient space to accommodate tesofensine within the binding site. Furthermore, strong ion density was observed at the previously reported Na2 and Cl sites⁴⁷, with a well-defined density for the surrounding residues. The Na2 site coordinates with the side chains of D421^TM8 and S422^TM8, as well as the carbonyl oxygens of G75^TM1, V78^TM1, and L418^TM8 (Supplementary Fig. 7a). The Cl site coordinates with the side chains of N82^TM1, Y102^TM2, Q317^TM6, S321^TM6, and S357^TM7 (Supplementary Fig. 7b). However, no density was observed at the Na1 site, likely due to tesofensine binding disrupting the stable coordination of Na1.

Fig. 2. Cryo-EM structure of DAT with tesofensine.

a Atomic model and cryo-EM map of the DAT bound to tesofensine. The tesofensine, sodium and chloride ions are represented as yellow, purple, and green spheres, respectively, and are labeled accordingly. Transmembrane helices are highlighted in red and labeled. b Cryo-EM density (blue mesh) along with the model of the tesofensine (yellow sticks) binding pocket. The cryo-EM density of tesofensine in the contour level of 0.3 in ChimeraX. c Concentration–inhibition curves of [³H]dopamine uptake by wild-type and indicated mutants of DAT with specified concentrations of tesofensine in the range of 10 pM to 100 μM. The IC₅₀ values of tesofensine are: WT, 9.12 nM; F326A, 816.5 nM; and S422A, 204.2 nM. Data are mean ± s.e.m. (n  =  3 biologically independent experiments). [³H]Dopamine uptake was normalized to uptake in the absence of tesofensine. Significant differences in IC₅₀ value were observed for F326A (P  =  0.0080) and S422A (P  < 0.0001) compared with WT. Two-sided unpaired t-test (threshold, α = 0.05). d 2D representation of protein-ligand interactions between tesofensine and DAT. The TMs are differentiated by background colors, aiding in the visualization of the interaction landscape. Residues conserved across MATs are indicated in red, while non-conserved residues are shown in black. e Structural comparison among DAT^teso, NET^teso (cyan), and SERT^teso (brown). Key residues are depicted as sticks and labelled, with the overlapping binding pocket highlighted by a blue dashed line.

Like cocaine, tesofensine also belongs to the tropane alkaloid family and features a tropane ring. Within the central binding pocket, the position of the tropane ring is nearly identical to that of cocaine⁵⁵ (Supplementary Fig. 7c). However, in cocaine, the tropane ring is connected to a phenyl group via an ester bond, allowing its phenyl group to extend deeper into the cavity between TM3 and TM8 compared to the dichlorophenyl group of tesofensine (Supplementary Fig. 7d). This difference in their binding poses weakens the π-π stacking interaction between the phenyl group and Y156^TM3, which may in part contribute to the 50-fold lower affinity of cocaine for DAT than tesofensine.

Tesofensine effectively targets MATs by inhibiting their reuptake ability. However, we identified variations in the residues involved in drug binding among different MATs (Fig. 2d). For instance, the residues F76 and F155 in DAT are substituted by F72 and Y151 in NET and Y95 and Y175 in SERT. Nevertheless, the variation in the hydroxyl group between phenylalanine and tyrosine does not significantly affect the ability of tesofensine to inhibit MATs. To investigate whether the binding mode of tesofensine is conserved in SERT and NET, we extended our structural analysis beyond DAT. We purified and determined the structures of human NET and SERT in complex with tesofensine (NET^teso and SERT^teso) at resolutions of 2.9 Å and 3.2 Å, respectively (Supplementary Fig. 8). Clear cryo-EM density for tesofensine enabled unambiguous modeling and comparative analysis (Supplementary Fig. 9). In all three MATs, tesofensine binds within the central binding cavity formed by TM1, TM3, TM6, and TM8, stabilizing an outward-facing conformation (Supplementary Fig. 9a, b). Superposition of the DAT^teso, NET^teso, and SERT^teso demonstrates that tesofensine occupies nearly identical positions with highly similar binding poses (Fig. 2e). Key interacting residues—such as D79, Y156, F320, and F326 in DAT, along with their counterparts in NET and SERT—adopt nearly identical side-chain conformations (Fig. 2e). This conserved interaction pattern explains tesofensine’s broad efficacy and highlighting the conserved inhibitory mechanism of TRIs across MATs. Notably, the ethoxymethyl group of tesofensine exhibits divergent orientations. In DAT^teso, it faces the extracellular side near TM2, whereas in the SERT^teso, it rotates toward the intracellular side and TM8. Intriguingly, in NET^teso, the ethoxymethyl group is observed in both extracellular- and intracellular-facing conformations, suggesting the coexistence of two binding modes (Supplementary Fig. 9c–e). This conformational flexibility likely arises from the absence of strong interactions between the ethoxymethyl group and surrounding residues, allowing it to sample multiple orientations. In contrast, the tropane ring and dichlorophenyl group maintain consistent, well-defined contacts across MATs, underscoring their central role in anchoring tesofensine within the binding pocket. Collectively, tesofensine engages all three transporters with similar affinities, and our comparative structural analysis highlights a shared inhibitory mechanism, along with subtle conformational adaptations to each MAT.

Inhibition of dasotraline

Dasotraline is a medication developed by Sunovion Pharma for the treatment of moderate to severe binge eating disorder (BED) and attention deficit hyperactivity disorder (ADHD)^56,57. It strongly inhibits DAT, NET, and SERT, with reported IC₅₀ values of 3 nM, 4 nM, and 15 nM, respectively⁵⁸. To elucidate the molecular mechanism underlying the binding and inhibition of monoamine transporters by dasotraline, we determined the complex structure of DAT with dasotraline (DAT^daso) at a resolution of 3.2 Å (Supplementary Fig. 3 and Supplementary Table 1). An extra density shaped like the number “7” was observed at the central binding pocket, aligning well with the molecular structure of dasotraline (Fig. 3a). The binding pocket of dasotraline is surrounded by TM1, TM3, TM6, and TM8, accessible from the extracellular side, while the intracellular gate is closed, indicating that dasotraline stabilizes DAT in an outward-facing conformation (Fig. 3b). Dasotraline features a tetrahydronaphthylamine ring system linked to an amine group and a dichlorophenyl group. The tetrahydronaphthylamine ring is positioned near the TM1 and TM6 helices, oriented towards the extracellular side, and is sandwiched between F76^TM1, F320^TM6, and F326^TM6, forming extensive hydrophobic interactions (Fig. 3c). The F326A mutation significantly diminishes the inhibitory potency of dasotraline by approximately 1500-fold (Fig. 3d), underscoring the pivotal role of F326 in dasotraline binding. Meanwhile, the positively charged amino group forms a stable salt bridge with the negatively charged carboxyl group of D79^TM1, with a distance of 4 Å (Fig. 3c). The dichlorophenyl group is deeply enclosed within the binding pocket, flanked by TM3 and TM8 helices, with the meta chlorine reaching a groove approximately 15 Å from the plane of the extracellular membrane (Fig. 3b, c). The dichlorophenyl group is enveloped by S149^TM3, V152^TM3, Y156^TM3, and S422^TM8, and its benzene ring engages in an edge-to-face π-π interaction with Y156^TM3, further stabilizing the binding of dasotraline (Fig. 3c). The S422A mutation led to a 1900-fold increase in the IC₅₀, highlighting its crucial role in dasotraline binding (Fig. 3d). We found that most of the residues involved in the dasotraline binding site are conserved among DAT, NET, and SERT (Fig. 3e), supporting the fact that it acts as a triple reuptake inhibitor.

Fig. 3. Delineation of dasotraline binding site.

a The dasotraline molecule is depicted in orange sticks, with its corresponding cryo-EM density rendered in blue mesh, presented in two distinct perspectives. The cryo-EM density of dasotraline in the contour level of 0.25 in ChimeraX. b Structure of DAT^daso in the outward-facing conformation. The overall structure is represented as blue cartoon, with dasotraline depicted in orange spheres. The extracellular cavity housing dasotraline is visualized with the electrostatic surface, and the depth of the cavity is labeled. c Close-up view of the dasotraline binding site. The dasotraline and key interacting residues are displayed as sticks and labeled. d Concentration–inhibition curves of [³H]dopamine uptake by wild-type and indicated mutants of DAT with specified concentrations of dasotraline in the range of 10 pM to 100 μM. The IC₅₀ values of dasotraline are: wild type (WT), 2.43 nM; F326A, 3736 nM; and S422A, 4701 nM. Data are mean ± s.e.m. (n  =  3 biologically independent experiments). [³H]Dopamine uptake was normalized to uptake in the absence of dasotraline. Significant differences in IC₅₀ value were observed for F326A (P  =  0.0222) and S422A (P  =  0.0405) compared with WT. Two-sided unpaired t-test (threshold, α = 0.05). e Ligplot analysis of dasotraline with DAT. The transmembrane membranes (TMs) are differentiated by distinct background colors to facilitate the visualization of the interaction landscape. Conserved residues across monoamine transporters (MATs) are marked in red, while non-conserved residues are indicated in black. f Structural comparison between the DAT^daso (blue) and SERT^sert (PDB:6AWO, gray). Arrows indicate the structural transitions, with critical residues shown as sticks and labelled.

Dasotraline is the stereoisomer of desmethylsertraline, which is the active metabolite of the commercially available selective serotonin reuptake inhibitor (SSRI) sertraline⁵⁹. However, sertraline demonstrates a higher selectivity for SERT over DAT and NET, with reported K_i (nmol/L) values of 2.8, 315 and 925, respectively⁶⁰. To elucidate the molecular basis of the distinct sensitivity of DAT to dasotraline and sertraline, we compared the structures of DAT^daso and SERT^sert using the scaffold domain as a reference⁶¹, yielding a root-mean-square deviation value of 1.5087 Å across 498 Cα atoms. A minor displacement of TM3 and TM6a in SERT compared with DAT leads to an overall alteration of the drug binding pockets in these two transporters (Fig. 3f). The dichlorophenyl group of sertraline penetrates the binding pocket approximately 1 Å deeper than that of dasotraline, causing clashes between S149^TM3 of DAT and the para-chloro group of sertraline (Fig. 3f). Additionally, we found that the salt bridge between the amino group of dasotraline and D79 may be stronger than that for sertraline because of the shorter distance and more favorable orientation of the amino group relative to D79 (Fig. 3f). These differences in interactions allow dasotraline to exhibit increased inhibitory potency for DAT. Although dasotraline and tesofensine are chemically distinct, both stabilize the DAT in its outward-facing conformation, and their binding poses and the residues involved in their binding are similar (Supplementary Fig. 9f). The amino groups of both drugs form electrostatic interactions with the carboxyl group of D79^TM1. These observations provide a structural explanation for the higher DAT selectivity of dasotraline over sertraline and highlight common features of TRI binding that underlie their inhibitory mechanisms.

Recognition of centanafadine

Centanafadine is a serotonin-norepinephrine-dopamine reuptake inhibitor (SNDRI) developed for the treatment of attention deficit hyperactivity disorder (ADHD) and is currently undergoing phase II and III clinical trials for various indications^45,46. It inhibits the reuptake of norepinephrine, dopamine, and serotonin at ratios of 1:6:14, with IC₅₀ values of 6 nM, 38 nM, and 83 nM, respectively⁶². The complex structure of DAT bound to centanafadine (DAT^cen) was determined at a resolution of 3.4 Å (Fig. 4a, Supplementary Fig. 4 and Supplementary Table 1). A rod-like density is observed at the halfway of the membrane (Fig. 4b). Distinct from outward-facing conformation stabilized by tesofensine and dasotraline, the centanafadine stabilized the DAT in the inward-facing conformation, with TM1a bending toward the membrane and the central binding site exposed to the intracellular side (Fig. 4a). Centanafadine consists of a naphthyl group and an azabicyclo group. The naphthyl group occupies the space between TM3 and TM8, while the azabicyclo group is situated closer to TM1 and TM6. S149^TM3, V152^TM3, and Y156^TM3 surround the naphthyl group, establishing extensive hydrophobic interactions (Fig. 4b, c). Mutating V152 to alanine resulted in a reduced apparent affinity for centanafadine, nearly doubling the IC₅₀ value compared with that of the DAT WT (Fig. 4d). G153^TM3 and G426^TM8, with their short side chains, allow the naphthyl group to penetrate deeply into the cavity (Fig. 4c). The azabicyclo group is encircled by D79^TM1, F320^TM6, and F326^TM6. Notably, the nitrogen atom within the azabicyclo moiety forms a hydrogen bond with the hydroxyl group of D79^TM1 at a distance of 2.7 Å (Fig. 4c). The F326A mutation markedly diminished the affinity for centanafadine, underscoring the importance of this interaction. All these structural observations are conserved in NET and SERT, supporting that it acts a triple reuptake inhibitor (Fig. 4e). Compared with other typical inhibitors, centanafadine is more compact and binds deeply within the binding pocket, completely overlapping the dopamine binding site visualized in the occluded state⁴⁷ (Fig. 4f). The naphthyl group of centanafadine occupies a position adjacent to the TM3 and TM8 helices, more so than the catechol group of dopamine. Additionally, the azabicyclo group of centanafadine engages in more extensive interactions with surrounding residues, likely enhancing its binding affinity for the DAT.

Fig. 4. Centanafadine binding pocket of DAT.

a Overall structure of DAT in complex with centanafadine depicted in cartoon representation with two different side views. Centanafadine is illustrated as deep green spheres, while the TMs are represented as lemon yellow cartoons. b The cryo-EM density of centanafadine, along with surrounding residues, is displayed as sticks. The cryo-EM density of centanafadine in the contour level of 0.25 in ChimeraX. c The binding pocket of centanafadine in DAT, viewed from the extracellular side. The key residues in the pocket are shown as sticks or spheres and labeled. d Concentration–inhibition curves of [³H]dopamine uptake by wild-type and indicated mutants of DAT with specified concentrations of centanafadine in the range of 0.1 nM to 100 μM. The IC₅₀ values of centanafadine are: WT, 116.1 nM; V152A, 246.1 nM; and F326A, 2094 nM. Data are mean ± s.e.m. (n  =  3 biologically independent experiments). [³H]Dopamine uptake was normalized to uptake in the absence of centanafadine. Significant differences in IC₅₀ value were observed for V152A (P  =  0.0290) and F326A (P  = 0.0043) compared with WT. Two-sided unpaired t-test (threshold, α = 0.05). e Schematics generated by Ligplot illustrate the interaction between centanafadine and DAT, with the green line denoting hydrogen bond. The TMs are distinguished by background colors, enhancing visualization of the interaction landscape. Conserved residues across MATs are indicated in red, while non-conserved residues are shown in black. f Structural comparison of the binding pocket between DAT^cen in the inward-facing conformation and DAT^DA in the occluded conformation. Centanafadine, dopamine, and key residues are shown as sticks and marked.

The ansofaxine binding site

Toludesvenlafaxine, commercially known as ansofaxine, is an approved SNDRI antidepressant in China and is currently in a phase III trial in the U.S. for treating depressive disorders^42,43,63. Ansofaxine is synthesized by appending a para-methylbenzoate group to the phenolic hydroxyl of desvenlafaxine, enhancing permeability and demonstrating high inhibitory potency against SERT, NET, and DAT with IC₅₀ values of 723 nM, 763 nM, and 491 nM, respectively⁶³. We determined the structure of the DAT in complex with ansofaxine (DAT^anso) at a resolution of 2.9 Å (Supplementary Fig. 5 and Supplementary Table 1), revealing a distinctive fork-like cryo-EM density representing ansofaxine (Fig. 5a). The structure of DAT^anso adopts an inward-facing conformation (Fig. 5b). Ansofaxine presents a dimethylamino and hydroxyl ring at its head, a phenyl group bridging the central structure, and a para-methylbenzoate group at its tail (Fig. 5a). Its binding pocket is primarily constituted by helices TM1, TM2, TM6, TM7, and TM8. The hydroxyl ring at the head is nestled beneath TM1, inducing an upward shift of TM1 and engaging in extensive hydrophobic interactions with F76^TM1, F114^TM2, L322^TM6, T349^TM7, T350^TM7 and N353^TM7 (Fig. 5c). The dimethylamino group extends toward the central cavity, in close proximity to D79^TM1. The central phenyl group and the terminal para-methylbenzoate moiety are situated over TM6, establishing further hydrophobic interactions with V328^TM6, L329^TM6, A331^TM6, F332^TM6, Y335^TM6, G425^TM8 and E428^TM8 (Fig. 5c). To validate the binding pocket of ansofaxine, we designed L329A and E428A mutations. [³H]Dopamine uptake experiments revealed a marked increase in the IC₅₀ of ansofaxine, from 730.4 nM in the DAT WT to 7110 nM for the L329A mutant and 5066 nM for the E428A mutant (Fig. 5d).

Fig. 5. The binding mode of ansofaxine in DAT.

a Molecular formula and cryo-EM density of ansofaxine. The cryo-EM density of ansofaxine in the contour level of 0.25 in ChimeraX. b DAT is stabilized in an inward-facing conformation by ansofaxine, which is embedded within a deep and negatively charged cavity. The surrounding TMs and the depth of the cavity are labeled. c Magnified view of the interaction network between ansofaxine and the TMs of DAT. Residues in close proximity to ansofaxine are depicted as violet sticks. d Concentration–inhibition curves of [³H]dopamine uptake by wild-type and indicated mutants of DAT with specified concentrations of ansofaxine in the range of 1 nM to 1 mM. The IC₅₀ values of ansofaxine are: WT, 730.4 nM; L329A, 7110 nM; and E428A, 5066 nM. Data are mean ± s.e.m. (n  =  3 biologically independent experiments). [³H]Dopamine uptake was normalized to uptake in the absence of ansofaxine. Significant differences in IC₅₀ value were observed for L329A (P  =  0.0316) and E428A (P  = 0.0283) compared with WT. Two-sided unpaired t-test (threshold, α = 0.05). e Schematic diagram of the interaction network between ansofaxine and its associated residues. The TMs that residues originate from are differentiated by distinct background colors. Conserved residues of MATs are marked in red, while non-conserved residues are presented in black. f Comparison of conformational changes and binding pocket between DAT^anso and DAT^cen in the inward-facing conformation. Ansofaxine and centanafadine are represented as violet and deep green sticks, respectively. The binding pockets are emphasized by red dashed lines.

The binding pocket for ansofaxine is situated near the intracellular side, bordered by highly conserved residues that contribute to its similar affinity for DAT, NET, and SERT (Fig. 5e). Notably, the binding sites of ansofaxine and centanafadine are completely non-overlapping (Fig. 5f). Centanafadine is located proximal to the central cavity, whereas ansofaxine is located closer to the intracellular cavity. Additionally, there are substantial structural differences in TM1a. The hydroxycyclohexyl group of ansofaxine induces a more pronounced extension and elevation of TM1a by approximately 10 Å, widening the cavity toward the cytoplasmic side (Fig. 5f). This observation suggests that in the inward-facing state, there may be distinct binding pockets that enable triple reuptake inhibition. Upon comparison of the DAT^anso with the apo-state DAT and DAT in complex with GBR12909 and dopamine (Supplementary Fig. 10), it was found that TM1a in DAT^anso exhibits the greatest upward tilt relative to the membrane plane. The ansofaxine binding site is also significantly distinct from that of GBR12909 (Supplementary Fig. 10b).

Antagonism of nefazodone

Nefazodone is an atypical antidepressant and exerts its effects as a potent inhibitor across various sites, including DAT, NET, and SERT, with reported equilibrium dissociation constants (K_D) values of 360, 360, and 200 nM, respectively^64,65. To explore the binding site of nefazodone, we determined the structure of nefazodone-bound DAT (DAT^nefa) at a resolution of 3.5 Å (Fig. 6a, Supplementary Fig. 6 and Supplementary Table 1). The cane-shaped density was consistent with that of nefazodone (Fig. 6b). The structure of DAT^nefa is captured in an inward-facing conformation (Fig. 6a). Nefazodone, a phenylpiperazine derivative, consists of four components: a chlorophenyl group, a piperazine group, a nitrogen-containing heterocycle, and a phenoxy group. Our cryo-EM map reveals that the resolution for nefazodone is modest. To further investigate this observation, we performed molecular dynamics simulations of the DAT^nefa complex and analyzed the root mean square deviation (RMSD) of nefazodone. Throughout the simulation, the overall RMSD of nefazodone remained consistently around 1.5 Å, indicating that the ligand maintains a stable conformation (Supplementary Fig. 11).

Fig. 6. Detailed representation of the binding site for nefazodone.

a Cartoon diagram of DAT in complex with nefazodone from side view. Nefazodone is rendered in blue spheres. b The cryo-EM density of nefazodone, along with surrounding residues, is displayed as sticks. Cryo-EM density of neafazodone in the contour level of 0.18 in ChimeraX. c Closer look into the interface interactions between nefazodone and TMs. Key residues involved in nefazodone coordination are displayed as sticks. d Concentration–inhibition curves of [³H]dopamine uptake by wild-type and indicated mutants of DAT with specified concentrations of nefazodone in the range of 0.1 nM to 100 μM. The IC₅₀ values of nefazodone are: WT, 1403 nM; F326A, 7518 nM; and L329A, 5409 nM. Data are mean ± s.e.m. (n  =  3 biologically independent experiments). [³H]Dopamine uptake was normalized to uptake in the absence of nefazodone. Significant differences in IC₅₀ value were observed for F326A (P  =  0.0139) and L329A (P  = 0.0429) compared with WT. Two-sided unpaired t-test (threshold, α = 0.05). e Schematic representation of nefazodone interaction in the binding pocket of DAT. The transmembrane segments from which the residues originate are distinguished by different background colors. Conserved residues of MATs are highlighted in red, whereas non-conserved residues are shown in black. f Superposition of DAT^nefa and DAT^anso in the inward-facing conformations, emphasizing the binding sites for nefazodone and ansofaxine in blue and red dashed lines, respectively.

The meta-chlorine of the chlorophenyl group extends deeply into the binding cavity, pointing between TM3 and TM8 (Fig. 6a). The phenyl group forms extensive hydrophobic interactions with S149^TM3, V152^TM3, Y156^TM3, and G426^TM8, and simultaneously engages in a T-shaped π-π stacking interaction with F326^TM6 (Fig. 6c). The piperazine group forms electrostatic interactions with the hydroxyl group of D79^TM1, thereby stabilizing the binding of nefazodone. The nitrogen-containing heterocycle is nestled among F76^TM1, F114^TM2, and L329^TM6, while the phenoxy group is strategically positioned at the cavity entrance, oriented between TM1 and TM7, and surrounded by S72^TM1, F332^TM6, and T349^TM7 (Fig. 6c). The phenyl group participates in π-π parallel stacking with F76^TM1 and vertical stacking with F332^TM6 (Fig. 6c). Trazodone, a selective serotonin reuptake inhibitor and a close analog of nefazodone, shows markedly reduced affinity for DAT, likely due to the lack of a phenoxy group that would otherwise facilitate π-π interactions. [³H]Dopamine uptake experiments revealed that the F326A and L329A mutants exhibited reduced sensitivity to nefazodone, with IC₅₀ values of 7.5 μM and 5.4 μM, respectively, approximately 5 and 4 times higher than that of DAT WT (Fig. 6d).

The binding pocket of nefazodone is formed by TM1-3 and TM6-8, featuring highly conserved interacting residues among MATs (Fig. 6e). This high degree of conservation is the basis for the similar affinities of nefazodone for DAT, NET, and SERT. Nefazodone, along with ansofaxine and centanafadine, stabilizes the DAT in an inward-facing conformation (Supplementary Fig. 12). Structural comparisons revealed notable differences between the binding pockets of nefazodone and ansofaxine, which exhibited an “X”-shaped crossover (Fig. 6e). The hydroxyl ring of ansofaxine extends beneath TM1a, whereas the chlorophenyl group of nefazodone is positioned between TM3 and TM8. Moreover, in the DAT^nefa complex, TM1a is closer to the membrane plane compared to the DAT^anso complex, with an inclination of 10 Å. These drug-bound structures demonstrate the diversity of drug-binding poses and establish a framework for how the intracellular cavity accommodates chemically distinct compounds.

Conformational transition of DAT

The structures of neurotransmitter sodium symporters, including DAT, norepinephrine transporter (NET), serotonin transporter (SERT), and glycine transporter 1 (GlyT1)^{47,48,66–68}, have been determined in various conformations. To understand the mechanistic differences and similarities in transport among these transporters, we compared them in each conformation (Supplementary Fig. 13). For the outward-facing conformation, the structures of DAT, NET, SERT, and GlyT1 are aligned using TM3, TM4, TM8, and TM9 as references (Supplementary Fig. 14a–c). We observed that the majority of the TM helices are similar. However, TM1b and TM6a in DAT are positioned closer to TM2 and TM7 than those in the other transporters (Supplementary Fig. 14d). Notably, the side chain orientation of F320 on TM6a is remarkably conserved across SLC6 transporters, closely aligning with its structural counterparts—F335 in SERT, F317 in NET, and Y316 in GlyT1. These residues are critical components of the ligand-binding pocket, playing a pivotal role in stabilizing the outward-facing conformation and modulating ligand interactions. On the intercellular side, the N-terminal loop region preceding the TMD has been well-resolved across all structures. A highly conserved network of interactions is observed, including the salt bridge between R60^TM1 and D436^TM10, the cation-π interaction between R60^TM1 and Y335^TM6, and the hydrophobic interactions among W63^TM1, Y335^TM6, and F69^TM1. (Supplementary Fig. 14e, f). These interactions are essential to stabilize TM1a, which is important for sealing off the intracellular cavity and stabilizing the transporter in the outward-facing conformation^69–71.

For the inward-facing conformations of these transporters, we used TM3, TM4, TM8, and TM9 as references for comparison (Supplementary Fig. 15a–c). We noticed that TM1a exhibited remarkable variation in conformation. Specifically, TM1a displayed more pronounced upward tilting in the structures of ibogaine-bound SERT, norepinephrine-bound NET, and ALX-5407-bound GlyT1, in contrast to that in DAT^nefa. We observed that both SERT and NET had a more pronounced outward tilt of TM1a compared to GlyT1. Additionally, while DAT and GlyT1 showed partial unwinding of TM5, SERT and NET retained helical integrity in this segment, which may relate to differential ligand properties and binding modes. Moreover, we observed that the benzene ring of F72 in NET is oriented toward the central pocket. However, the benzene rings of F76 in DAT, Y95 in SERT, and Y62 in GlyT1 are directed toward the cytoplasmic side, facilitating a larger opening on the intracellular side (Supplementary Fig. 15d). On the extracellular side, the salt bridge formed between D476^TM10 and R85^TM1 is conserved among NSS members, which is critical for sealing the extracellular cavity⁷⁰ (Supplementary Fig. 15e). Moreover, structural comparison of the C-terminal tail of NSS members revealed that the last residues of DAT and NET can be clearly resolved with highly similar structural arrangements (Supplementary Fig. 15f). In contrast, SERT, NET, and GlyT1 exhibit varying conformations at their C-terminal tails, which presumably allows these transporters to interact with distinct regulatory proteins.

Discussion

Tricyclic antidepressants, the first generation of antidepressants, were classified as triple reuptake inhibitors due to their effects on serotonin, norepinephrine, and weak dopamine pathways. However, their clinical utility is limited by non-selective interactions with muscarinic, histaminergic, and adrenergic receptors, resulting in adverse effects such as cognitive impairment, sedation, and cardiovascular toxicity^18,21,22. In contrast, the TRIs described herein produce nearly equivalent inhibition of MATs with several potential advantages, including minimizing the risk of adverse events or toxic reactions associated with polypharmacy, shortening the latency period for therapeutic effects, and mitigating side effects caused by serotonin activation by enhancing dopaminergic activity^72,73. Therefore, TRIs represent a promising class of potential antidepressants, with applications extending to the treatment of ADHD and obesity^{39,43,46,52,74}. However, most TRIs are still in clinical trials, and the extent to which they will advance into therapeutic application is yet to be determined. Elucidating the details of the binding modes of TRIs with MATs is important. Our study has systematically revealed that TRIs bind to conserved pockets in different functional states. These structural insights could provide a basis for the rational optimization of existing chemical scaffolds when these drug candidates encounter limitations or fail during development.

In this study, we have elucidated DAT bound to five distinct TRIs under clinical investigation, each with a distinct core chemical structure. The high-resolution structures, spanning from 2.8 to 3.5 Å, have facilitated the identification of diverse pharmacophores and the elucidation of their specific mechanisms of action. The TRIs stabilize DAT in two key functional states: the inward-facing and outward-facing conformations, providing critical insights into their varied binding profiles. Although tesofensine and dasotraline are chemically distinct compounds, they both stabilize the DAT in its outward-facing conformation, with their binding pockets displaying striking structural similarities. Both drugs engage in extensive hydrophobic interactions with TM1, TM3, TM6, and TM8, forming a stable salt bridge with D79 and a π-π stacking interaction with Y156—key elements that enhance drug affinity. While dasotraline and sertraline share structural similarities, sertraline’s higher SERT selectivity correlates with its deeper penetration into the SERT binding pocket, as evidenced by different interaction patterns, TM3/TM6a displacement and steric clashes in DAT. Structural comparison of DAT^teso, NET^teso, and SERT^teso reveals a conserved binding mode that stabilizes the transporters in an outward-facing conformation through conserved interactions with homologous residues in the central substrate-binding cavity. These results provide mechanistic insights into the conserved inhibitory action of TRIs across the MAT family.

Ansofaxine, centanafadine, and nefazodone stabilize the DAT in the inward-facing conformation. Their binding pockets display considerable structural diversity, reflecting the flexibility and versatility of the intracellular binding pocket. Nefazodone and trazodone share structural similarities and both act on multiple targets to exert antidepressant effects⁷⁵. However, nefazodone inhibits SERT activity in a non-competitive manner, while trazodone demonstrates mixed-competitive inhibition of SERT⁷⁵. Such differences in inhibition pattern may arise from their distinct binding modes. Trazodone can bind to SERT simultaneously with serotonin, potentially occupying the S2 site of SERT to exert a mixed-competitive effect⁷⁵. Regarding the nefazodone, our study reveals that nefazodone binds to the substrate-binding pocket and stabilizes the transporter in an inward-facing conformation, thus preventing substrate binding and exhibiting a non-competitive inhibition mode. Furthermore, ligands that bind to the site distinct from substrate binding pocket and regulate protein function indirectly are allosteric modulators⁷⁶. Among the structures we elucidated, TRIs either coincide with the orthosteric site or are adjacent to the substrate binding site, and thus do not act as allosteric modulators. Centanafadine is positioned deeper within the central cavity, while ansofaxine extends beneath TM1a. Notably, the binding of ansofaxine elicits a distinct interaction pattern. The pronounced upward shift of TM1a alters the shape of intracellular binding pocket, potentially affecting the binding kinetics. In the recent study of NET, the G425 of NET is supposed to be critical for drugs discriminate between NET or DAT and SERT, as the equivalent position in DAT/NET and SERT is occupied G422 and A441, respectively⁶⁶. For the TRIs bound in the intracellular cavity of DAT, we found that they are positioned either above or below this site, or near this residue with a thin group, to achieve broader inhibition of MATs.

Taken together, our intricate structural insights provide a comprehensive blueprint of how chemically distinct triple reuptake inhibitors associate with either the extracellular cavity or the intracellular pocket across various MATs, thereby inhibiting transporter activity with broader selectivity. This information is crucial for the development of the next generation of antidepressants, which are expected to be more potent and have fewer side effects. While this study reveals distinct patterns of conformational stabilization and structural determinants of TRI binding to MATs, the functional consequences of these conformational differences, including their potential effects on neurotransmitter dynamics, remain to be determined. Future investigations will be essential to establish how these structural states translate into distinct regulatory and therapeutic outcomes.

Methods

Protein expression, purification and nanodisc reconstitution

The plasmids encoding DAT and NET were described previously^47,66. Expression and purification of the corresponding proteins followed similar procedures. Human full-length SERT (SLC6A4; UniProt ID: P31645) cDNA was obtained from a cDNA library and cloned into a modified pEG BacMam vector. This vector incorporates a C-terminal tag sequence comprising a 3C protease cleavage site, green fluorescent protein (GFP), and a Twin-Strep II affinity tag. Subsequently, recombinant baculoviruses were generated using the Bac-to-Bac system, and high-titer virus stocks were produced by propagating the viruses in Sf9 insect cells. For protein expression, HEK293F cells were cultured in a shaker at 37 °C, under 5% CO₂. When the cell density reached approximately 2.5 × 10⁶ cells per mL, 2% (v/v) P2 recombinant baculovirus was to the cell culture, supplemented with 1% (v/v) fetal bovine serum. After culturing for 12 hours, cells were treated with 10 mM sodium butyrate and further incubated at 30 °C with 5% CO₂ for 48 hours. Cells were then harvested at 3000 rpm for 3 minutes, flash-frozen, and stored at −80 °C for later purification.

Cells were resuspended in lysis buffer A (20 mM HEPES pH 7.5 (High Purity Grade, JS0164, JSENB), 150 mM NaCl, 15 mM β-ME) supplemented with 2 μg/mL aprotinin, 1.4 μg/mL leupeptin, 0.5 μg/mL pepstatin A (MedChemExpress, USA). The cell membranes were pelleted by ultracentrifugation at 100,000 × g for 40 min and then were resuspended in buffer A supplemented with an additional 2 mg/mL iodoacetamide rotating for 30 minutes. Protein was then extracted with buffer B (buffer A supplemented with 1% (w/v) Lauryl maltose neopentyl glycol (LMNG) and 0.15% (w/v) cholesteryl hemisuccinate (CHS, Anatrace)) by gentle agitation for 2 h at 4 °C. After extraction, the supernatant was collected after ultracentrifugation at 100,000 × g for 1 h and then passed through Strep-Tactin resin (Smart-Lifesciences, China) at 4 °C in a gravity column. The column was washed with 10 column volumes of buffer C (buffer A supplemented with 0.01% LMNG and 0.0015% CHS). Protein was eluted with buffer D (buffer A plus 4 mM d-desthiobiotin (1169249, Leyan), 0.01% (w/v) LMNG and 0.0015% CHS) and concentrated with a 50-kDa cut-off concentrator (Merck Millipore, Germany). The protein sample was further purified by size-exclusion chromatography on the Superose 6 increase10/300 GL column (GE Healthcare) pre-equilibrated with buffer C. The peak fraction of protein was collected and concentrated to an appropriate concentration of 1 mg/ml for nanodisc reconstitution. To capture the interactions with various ligands, ansofaxine, centanafadine, tesofensine, nefazodone, and dasotraline (MedChemExpress, USA) were introduced during both the cell collection and the purification process. The concentrations of these ligands were meticulously adjusted to 20 μM for ansofaxine and 10 μM for each of the other ligands.

The nanodisc reconstitution was performed using membrane scaffold protein MSP1D1E3 and brain polar lipid. MSP1D1E3 was purified via a Ni-NTA resin, followed by extensive dialysis in TBS buffer (20 mM Tris pH 8.0, 150 mM NaCl). Brain polar lipid (BPL, Avanti) and cholesterol at a ratio 3:1 (w/w) in chloroform and desiccated under vacuum overnight. The lipid mixture was then solubilized in HBS buffer (20 mM HEPES, 150 mM NaCl, pH 7.5) to a final concentration of 25 mg/mL and incubated with 2% (w/v) LMNG at room temperature for 30 minutes. Purified protein was combined with MSP1D1E3 and the lipid mixture at a molar ratio of 1:5:200, and the mixture was gently rotated for 1 h at 4 °C. To remove detergent, Bio-Beads (SM2, BioRad) were then added to the mixture in a final concentration of 600 mg/mL and rotated overnight at 4 °C. After centrifugation to remove the Bio-Beads and excess lipids, the reconstituted mixtures were purified with StrepII-Tactin resin. The elute was treated with HRV-3C protease to cleave GFP, and HRV-3C protease was subsequently removed using a Ni-NTA column. The flow-through sample was concentrated to 1 mL using a 50-kDa Millipore tube and subsequently separated on a Superose 6 Increase 10/300 GL column under the conditions of buffer A. The peak fractions were collected and concentrated to 10–20 mg/mL before cryo-EM sample preparation. The peak fractions were also evaluated via 12% SDS-PAGE with prestained standard protein marker (P9006, New Cell & Molecular Biotech).

Cryo-EM sample preparation and data collection

For the preparation of samples with bound ligands, higher concentrations of the ligands—1 mM for ansofaxine, centanafadine, and tesofensine, 200 μM for nefazodone, and 400 μM for dasotraline—were incorporated into the final, concentrated protein sample prior to cryo-sample preparation. Cryo-EM grids were prepared by applying 2.5 μL of purified protein onto a freshly glow-discharged Quantifoil holey carbon grid (R1.2/1.3, Cu, 300 mesh). The grids were then blotted with filter paper for 4.5 s at 4 °C under 100% humidity, followed by plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific, USA). The grids were mounted on a 300 kV Titan Krios transmission electron microscope, equipped with a Gatan K3 Summit detector and GIF Quantum energy filter (with 20 eV slit for all ligands). Data acquisition was performed using EPU software, with the microscope operated at a calibrated magnification of 105,000x in super-resolution mode, resulting in 0.85 Å/pixel, with a defocus range of −1.0 to −2.0 μm. For the DAT^nefa dataset, the cryo-EM data of DAT^nefa were dose-fractionated across 40 frames, accumulating to a total dose of ~50 e⁻ Å⁻². For the remaining datasets, each movie stack was dose-fractioned into 32 frames, yielding a total accumulated dose of ~60 e⁻ Å⁻².

Cryo-EM data processing

For all datasets, image processing and reconstruction were performed using CryoSPARC⁵⁰. The processing workflow for each of the cryo-EM structures is summarized in Supplementary Figs. 2,3,4,5,6, and 8. Movies were pre-processed with Patch Motion Correction and Patch CTF Estimation. Particle picking was performed using the blob picker and Topaz picker (after blob picking on a subset of micrographs), followed by extraction with a box size of 256 pixels. Class averages, obtained from several rounds of 2D classification were used to select a subset of particles for reference-free ab-initio reconstruction. All extracted particles were then refined through heterogeneous refinement using the ab-initio model. Finally, map density was improved through multiple rounds of non-uniform refinement, local refinement and 3D classification without alignment, resulting in a 3D reconstruction at resolutions of 2.9 Å, 3.4 Å, 3.4 Å, 3.2 Å, 2.8 Å, 2.9 Å and 3.2 Å for DAT^anso, DAT^cen, DAT^nefa, DAT^daso, DAT^teso, NET^teso and SERT^teso, respectively.

Model building and refinement

Model building was carried out using Coot⁷⁷ and refined with real-space refinement in Phenix⁷⁸. The structural models for DAT^anso, DAT^cen and DAT^nefa were constructed based on the initial model of DAT^apo (PDB: 8Y2C). Similarly, the structural models for DAT^daso and DAT^teso were built using DAT^MPH (PDB: 8Y2G) as a template, NET^teso was modeled on NET^atomoxetine (PDB: 8Z1L), and SERT^teso was based on SERT^5-HT (PDB: 7LIA). For the ligands, two-dimensional (2D) structures were downloaded from PubChem in the SDF format. Three-dimensional (3D) models and their geometric constraints were generated using the eLBOW module in PHENIX. The initial models were fitted into the corresponding cryo-EM density maps by UCSF Chimera⁷⁹. The ligand molecules were then docked into the EM maps and optimized according to the corresponding density. All residues were manually inspected and adjusted to fit the maps in Coot, followed by multiple rounds of real-space refinement in PHENIX. Model validation was conducted using the comprehensive validation (cryo-EM) in PHENIX.

All figures were prepared with PyMOL, UCSF Chimera⁷⁹ or LigPlot+⁸⁰.

[³H] Dopamine transport assay

[³H]Dopamine transport assay was conducted according to previously published experimental methods⁴⁷. Briefly, HEK-293F cells were infected with different viruses and collected for experiment after 18 h expression. Cells were washed once using the uptake buffer (5 mM Tris base, 7.5 mM HEPES, 120 mM NaCl, 5.4 mM KCl, 1.2 mM CaCl₂, 1.2 mM MgSO₄,1 mM ascorbic acid, and 5 mM glucose at pH 7.4), and were subsequently incubated with various concentrations of inhibitor for 10 minutes at room temperature. Then the mixture of 5 nM [³H]Dopamine and 1 μM unlabeled dopamine was added to initiate reaction. The uptake was terminated at 2 min by adding cold (4 °C) uptake buffer and rapid centrifugation at 17000 x g for 10 s. After a final wash with cold uptake buffer, cells were lysed in 1% Triton X-100 followed by incubation with 2 mL of scintillating agent (Optiphase Supermix, PerkinElmer) for 5 min. Then, samples were counted using Hidex 300 SL liquid scintillation counter (Hidex, Finland). The uptake activity of uninfected cells represented as background value. The specific uptake value was calculated by subtracting the background from total uptake, followed by normalizing to the value measured with no inhibitor. The assays were three biologically independent experiments. IC₅₀ values were determined by fitting dose-response curves and calculated using the equation Y = 100/(1 + 10^(X-LogIC50)) in GraphPad Prism9. The primer sequences for the mutants used in this study are provided in Supplementary Table 2.

Molecular dynamics simulations

To verify the binding mode of nefazodone with DAT, we conducted molecular dynamics (MD) simulations on the DAT^nefa complex structure. To maintain the integrity of the ligand-binding pocket and prevent abnormal lipid penetration during the simulations, we completed the loop between TM4 and TM5. The initial simulation system was constructed via CHARMM-GUI⁸¹, comprising the DAT embedded in a POPC lipid bilayer, TIP3P water⁸² and ligand molecules. Physiological ionic strength was maintained using 150 mM NaCl, with charge neutralization achieved through the placement of ions. The system consists of a DAT^nefa complex, 123 POPC molecules, 80 ions, and 15,020 water molecules, with a total of 70,211 atoms. Force field parameters were assigned as follows: Amber ff14SB for proteins and lipids, GAFF2 for ligands^81,83. Energy minimization and pre-equilibration steps adhered to the standard CHARMM-GUI protocols. Three independent 100 ns production simulations were conducted under NPT ensemble using GROMACS 2021.6. The simulations were conducted with temperature coupling set at 303.15 K using a v-rescale thermostat⁸⁴ and pressure regulation maintained at 1 bar via the C-rescale barostat⁸⁵. A 2-fs integration time step was applied with the LINCS constraint algorithm. Long-range electrostatics were managed using the Particle Mesh Ewald (PME) method with a real-space cutoff of 0.9 nm, while van der Waals interactions employed a smooth switching function with a cutoff range of 0.9 to 1.0 nm. The calculation of RMSD for proteins and ligands was performed using VMD⁸⁶.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Y.Z. and G.W. conceived and supervised the project. Y.L. carried out molecular cloning and made all the mutation constructs. Y.L. and N.L. expressed, purified the protein, and prepared the sample for cryo-EM study. Y.M. performed functional assay. J.Z. and R.L. carried out cryo-EM data collection. Y.L. processed the cryo-EM data. Y.L. and Y.M. built and refined the atomic model. Q.B. performed the molecular dynamics simulations. Y.L., Y.M., and Q.B. analyzed the structures and prepared the figures. Y.Z., Y.L. and Y.M. wrote and revised the manuscript.

Peer review

Peer review information

Nature Communications thanks Harald Sitte and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Unless otherwise stated, all data supporting the results of this study can be found in the article, supplementary, and source data files. The initial model of DAT^apo is derived from Protein Data Bank under accession code 8Y2C. The three-dimensional cryo-EM density maps of the DAT^daso, DAT^teso, DAT^cen, DAT^anso, DAT^nefa, NET^teso and SERT^teso have been deposited in the Electron Microscopy Data Bank under the accession code EMD-61180, EMD-61181, EMD-61182, EMD-61184, EMD-61185, EMD-65400 and EMD-65401. The coordinates for the DAT^daso, DAT^teso, DAT^cen, DAT^anso, DAT^nefa, NET^teso and SERT^teso have been deposited in Protein Data Bank under accession codes 9J6R, 9J6S, 9J6T, 9J6U, 9J6V, 9VWR and 9VWS. The molecular dynamics simulation dataset has been uploaded to Zenodo [https://zenodo.org/records/17451922]. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-66670-3.