Main

Chemical neurotransmission is initiated by Ca2+-induced release of neurotransmitters into the synaptic cleft1. Upon release into the synaptic cleft, neurotransmitters such as glutamate, dopamine, noradrenaline, serotonin, glycine and GABA (γ-aminobutyric acid) activate G-protein-coupled receptors and ligand-gated ion channels, resulting in excitatory or inhibitory postsynaptic signalling cascades and currents1,2,3. The widespread and critical roles of neurotransmitters in both central and peripheral nervous systems necessitate a requirement for strict spatiotemporal control of their levels at neural synapses. The primary mode of neurotransmitter clearance from the synaptic cleft is through secondary active transporters localized in presynaptic cells and glial cells that harness ionic gradients, across the cell membrane, to drive the uphill transport of neurotransmitters4. This symport process requires both Na+ and Cl− ions5, which has led to the solute carrier 6 (SLC6) family of secondary transporters3 being referred to as neurotransmitter sodium symporters (NSSs)2.

Dysregulation of NSS function is associated with several debilitating disorders that include depression2, attention deficit hyperactivity disorder6, orthostatic intolerance7, epilepsy8, Parkinson’s disease2 and infantile parkinsonism dystonia9. NSSs are also the primary targets of antidepressants, drugs to treat neuropathic pain, attention deficit hyperactivity disorder, anxiety and of habit-forming substances of abuse such as cocaine and amphetamines3. Development of antidepressants had a serendipitous beginning in the 1950s10, followed by the discovery that the tricyclic antidepressant (TCA) imipramine inhibits noradrenaline reuptake in tissues11. Numerous variants of imipramine, and the subsequent discovery of selective serotonin reuptake inhibitors, have revolutionized antidepressant treatment12,13. To date, inhibition of neurotransmitter uptake remains the most widely used strategy for antidepressant therapy12, despite numerous side effects14.

Gains in our understanding of the molecular mechanisms underlying sodium-coupled transport have benefited from the structures of multiple conformations of LeuT15,16,17, a bacterial sodium-coupled amino acid transporter with ∼20% sequence identity to the eukaryotic NSSs. Models of eukaryotic NSSs based on LeuT have provided valuable insights into substrate and ion specificities, pharmacology and transport mechanisms in NSS members18,19. However, bacterial NSS models fall short of answering questions concerning the elements of NSS structure and function, including the local structure of NSSs in regions that are unrelated to LeuT in amino acid sequence, the determinants of substrate selectivity and the atomic-level details of transport inhibition by antidepressants and addictive compounds. Moreover, there is no understanding, at the level of three-dimensional structure, of the role of lipids and post-translational modifications in NSS structure and mechanism.

Here we present a 3.0 Å X-ray crystal structure of the Drosophila melanogaster dopamine transporter (DAT)20 in complex with the TCA nortriptyline. The Drosophila DAT has greater than 50% sequence identity with its mammalian counterparts and harbours a pharmacological profile that is a hybrid of the mammalian DATs, noradrenaline transporters (NETs) and serotonin transporters (SERTs), making it a powerful vehicle to study NSS pharmacology and substrate specificity20. The DAT structure reveals atomic details of TCA recognition, novel structural elements of NSS protein architecture and suggests a role for cholesterol in the allosteric control of transport in eukaryotic NSS members.

Thermostabilization and crystallization

Wild-type Drosophila DAT is labile, loses ligand-binding activity upon detergent extraction from the cellular membranes and is refractory to crystallization. To stabilize DAT for functional characterization, antibody generation and crystallization, we screened single point mutants for ligand-binding activity at increased temperatures21, ultimately combining five mutations into the construct used for crystallization and structure determination (DATcryst; Supplementary Fig. 1). Purified DATcryst binds to the high-affinity inhibitor nisoxetine with a dissociation constant (Kd) of 29 nM (Supplementary Fig. 2a), and the TCA nortriptyline exhibits an inhibition constant (Ki) of 156 nM (Supplementary Fig. 2b). Unfortunately we were unable to measure the binding of nortriptyline to wild-type DAT because of its instability. Nortriptyline has a Ki of 18 nM at human SERT and 4.4 nM at human NET22, values that are ∼9-fold and ∼35-fold lower than that for Drosophila DATcryst. In dopamine-uptake measurements with the wild-type DAT and amitriptyline, a precursor of nortriptyline, transport is inhibited with a Ki of 30 nM20, whereas the DATcryst construct is inactive in transport (Supplementary Fig. 2c, d). Crystallization was further enhanced by the use of a complex with a Fab, resulting in crystals of a DATcryst–Fab complex that diffract X-rays to 3.0 Å resolution.

Architecture of DAT

The structure of Drosophila DATcryst bound to nortriptyline exhibits an outward-open conformation whereby the antidepressant is bound in a cavity halfway across the membrane bilayer and accessible to solvent from only the extracellular side of the membrane (Fig. 1). The transporter displays an overall LeuT-like fold with 12 transmembrane helices (TMs) in which helices 1–5 and 6–10 are related by inherent pseudosymmetry, akin to LeuT15 (Supplementary Fig. 3). Residues in TM1 and TM6 make numerous interactions with the ligand and ions via non helical, hinge-like regions at the approximate mid-points of these TMs, connecting the bonding networks of all three ions with the inhibitor. Residues at the bend in TM3 contribute to the hydrophobic pocket that cradles the tricyclic moiety of the ligand, which lays approximately perpendicular to the TMs, mimicking a wedge separating the jaws of a vice. One cholesterol molecule is located in a groove between TM5 and TM7 and poised to modulate the movement of TM1a that occurs during the transport cycle (Fig. 1a)17.

Figure 1: Architecture of Drosophila DATcryst.
figure 1

a, Structure of DATcryst viewed parallel to membrane. Nortriptyline, sodium ions, a chloride ion and a cholesterol molecule are shown in sphere representation in magenta, purple, green and yellow, respectively. b, View of DATcryst from the extracellular face. c, Surface representation showing that ligand and ion binding sites are accessible from the extracellular vestibule. Nortriptyline and TMs 1, 3, 6 and 8 are coloured as in a.

PowerPoint slide

Full size image

The primary binding site accommodates nortriptyline but cannot adopt the subsequent helical movements of TMs 1b and 6a required to form the occluded state. Using LeuT for comparison, the occluded state of LeuT is formed in the presence of sodium and leucine substrate, but not in the presence of tryptophan, which binds to the primary site, comparable to the TCA in the context of DATcryst. We propose that both tryptophan and TCA stabilize the outward-open conformations of LeuT and DATcryst, respectively, by targeting the primary binding site and sterically blocking the extracellular domains of the transporter, preventing the extracellular gate from closing and thus acting by way of a foot-in-the-door mechanism (Supplementary Fig. 4a, b and Supplementary Table 2).

Whereas the core of DATcryst closely resembles that of LeuT, the periphery of DATcryst exhibits several features distinct from LeuT and important for neurotransmitter transport and cellular localization. In TM12, a kink in the centre at Pro 572 causes the second half of the helix to turn away from the transporter, indicating that the dimerization interface of LeuT is not the same as potential oligomerization interfaces of eukaryotic NSSs (Fig. 1a, b and Supplementary Fig. 5). Although previous studies indicate that NSSs oligomerize23,24, DATcryst is monomeric in detergent micelles and in the crystal lattice (Supplementary Fig. 6), thus suggesting that a membrane bilayer or additional molecules may be required for NSS assembly. The variable extracellular loop 2 (EL2) region has numerous predicted N-linked glycosylation sites25 and one disulphide bond26, modifications that have critical roles in proper trafficking of NSSs to the plasma membrane3. The strictly conserved disulphide linkage26 was observed in the structure between two conserved cysteines, Cys 148 and Cys 157 (Supplementary Fig. 7). In the crystal, EL2 has a central role in lattice contacts, packing against a neighbouring Fab with an 870 Å2 interface (Supplementary Fig. 6). Because 43 residues were deleted from EL2 in the DATcryst construct, further studies are required to determine the role of the full-length EL2 in transporter structure and function (Supplementary Fig. 8a). Together with EL2, EL4 harbours a Zn2+-binding site in mammalian DATs that modulates transport27. The equivalent residues in Drosophila DATcryst are within a Cα–Cα distance of 10 Å, but because their identities are Glu 161, Leu 374 and Ala 395, they do not form a high-affinity Zn2+-binding site in DATcryst.

TCA-binding site

Unambiguous density for nortriptyline in DATcryst was observed in the primary site, approximately halfway across the membrane bilayer (Fig. 2a). In accordance with previous chimaeric studies, swapping TM regions between NET and DAT28, the drug-binding site is surrounded primarily by helices 1, 3, 6 and 8 in a region equivalent to the substrate-binding pocket of LeuT15, and in close proximity to the densities for sodium and chloride ions (Fig. 2a). The dibenzocycloheptene ring of nortriptyline is oriented as a saddle, curving around the central region of TM3 and engaging in hydrophobic interactions with Val 120, Tyr 124 and Ala 117 (Fig. 2b). Val 120 is extensively conserved and faces the cycloheptene ring (Supplementary Fig. 4a), and replacement of the corresponding Ile 172 in human SERT with larger substitutions such as methionine markedly reduce affinity towards most NSS inhibitors19. This location was previously found in human SERT to be protected from crosslinking agents in the presence of inhibitor or substrate29. Phe 325 in TM6b forms an edge-to-face aromatic interaction with one of the benzyl groups of nortriptyline. Residues Gly 425 (TM8) and Ala 479 (TM10) also interact with the tricyclic group of the drug. The N-methylpropylamine group of the drug extends across the width of the drug-binding site and prevents TMs 1b and 6a from closing the extracellular gate ‘above’ the drug. The amine group forms a hydrogen bond with the main-chain carbonyl of Phe 43 and a cation-π interaction with the side chain of Phe 43 (Fig. 2b). Interestingly, residues equivalent to Val 120 and Phe 43 (Ile 172 and Tyr 95) in SERT are necessary for interactions with antidepressants30.

Figure 2: Antidepressant-binding site.
figure 2

a, Overall view of the nortriptyline-bound DATcryst. Fo – Fc densities (blue mesh) for drug and ions are contoured at σ levels of 2.5 and 3.0, respectively. b, Close-up view of the drug binding pocket. Na+ and Cl− ions are shown as spheres. Nortriptyline is represented as sticks (magenta). The amino group of nortriptyline is 2.7 Å from the carbonyl oxygen of Phe 43 (TM1a) and the N-methyl group of nortriptyline is 3.1 Å from the carbonyl oxygen of Phe 319. Residues lining the drug-binding pocket with interfacial areas greater than 10 Å2 are represented as sticks. c, Comparison of the drug or substrate binding pocket of DATcryst with that of LeuT (PDB code 2A65). The distance between the carboxylate group of leucine and Tyr 108 (spheres) is 2.7 Å in the occluded state (2A65) and 5.1 Å in the inhibitor-bound state (3F3A) of LeuT, whereas the equivalent interaction in DATcryst between Asp 46 and Tyr 124 is 3.1 Å.

PowerPoint slide

Full size image

The biogenic amine transporters harbour a crucial aspartate residue in TM1 and in the DATcryst structure we see how Asp 46 substitutes for the absence of the carboxylate group in biogenic amines as compared to amino acid substrates transported by LeuT and the GABA and glycine transporters (Supplementary Fig. 8b)15. The side chain of Asp 46 forms a hydrogen bond with the hydroxyl of Tyr 124, which is equivalent to the Tyr 108 residue in LeuT that has a role in substrate recognition (Fig. 2c)31. Mutations at this aspartate result in substantial losses in transport activity and reduced binding affinities for cocaine32. Ser 421 (TM8), which coordinates a sodium ion at site 2 (Na2), is within 3.5 Å of the propylamine group of the TCA and also forms a hydrogen bond with the carbonyl of Phe 43. Ser 421 therefore participates in a network of hydrogen bonds that interconnects nortriptyline with the Na2 site and was also found to be crucial for high-affinity recognition of antidepressants by human SERT33.

The N-methyl group of nortriptyline is 3.1 Å away from the main-chain carbonyl of Phe 319 and sterically prevents Phe 319 and TM6a from closing the extracellular gate above the drug, thereby stabilizing the outward-open state of the transporter. Phe 319 is the equivalent of Phe 253 in LeuT, which gates the substrate-binding pocket (Fig. 2c)15,16,17. The relative position of Phe 319 is markedly different from Phe 253 in the substrate-bound, occluded structure of LeuT, and instead resembles the positions of Phe 253 in the substrate-free and inhibitor-bound structures. To address the question of whether nortriptyline could bind to DATcryst in a LeuT-like, occluded conformation, we superimposed DATcryst onto the occluded state of LeuT and found that Phe 319 and Phe 325 would clash with the dibenzocycloheptene ring of the TCA (Fig. 2c and Supplementary Fig. 4c). Identification of nortriptyline bound in the substrate-binding pocket of DATcryst provides the first structural evidence that TCA antidepressants inhibit neurotransmitter transporters by preventing substrate binding and stabilizing the outward-open conformation18,19,34. The DATcryst–nortriptyline complex, together with the LeuBAT–antidepressant complexes (Wang, H. et al., unpublished observations), demonstrate conclusively that antidepressants inhibit NSSs by acting at the primary or S1 site, in stark contrast to how TCAs inhibit LeuT via a non-competitive mechanism35 by binding within the extracellular vestibule35,36,37.

Ion-binding sites

Locations of ions essential for transport could be identified in DATcryst with electron densities (>4.0σ) at three locations near the non-helical hinge-like regions of TMs 1 and 6, and close to the TCA. Densities at the two sites coincided exactly with Na1 and Na2 sites identified in LeuT (Fig. 3a, b)15. A chloride ion was positioned at the third position of high omit density nestled in between TMs 2, 6 and 7 and close to Na1 (Fig. 3a). Placing ions in the omit densities during model building led to a concomitant loss of Fo – Fc density during refinement. The atomic displacement factors of the ions matched the B-values of surrounding atoms. The sodium at site 1 is located ∼5.2 Å away from the amino group of nortriptyline and is coordinated with an octahedral geometry by side-chain oxygens of Asn 49, Ser 320 and Asn 352, and main-chain carbonyls of Ala 44 and Ser 320 (Fig. 3a and Supplementary Fig. 8b). Interestingly, the sodium at Na1 is also coordinated by one water molecule which in turn is within hydrogen-bonding distance to Asp 46, thus showing that the Asp in TM1 indirectly participates in the sodium ion coordination. The mean ion coordinating distances (2.6 Å) at this site are longer than the distances (2.42 Å) reported for Na+ ions in solution but shorter than the distances reported for K+ ions (2.84 Å; Supplementary Table 3)38.

Figure 3: Ion-binding sites.
figure 3

a, Na1 and chloride ion binding sites. Na+ is purple and Cl− is green and both are modelled as spheres. b, Coordination at the Na2 site is trigonal bipyramidal with the water molecule (w, red sphere) 3.3 Å from the sodium ion. Distances are in ängstroms for residues that are in the coordination sphere and interactions are shown by dashed lines. Residues are coloured according to their respective TMs.

PowerPoint slide

Full size image

The chloride ion is located 5.0 Å away from the Na1 site at a position previously identified by computational and mutational studies based on LeuT39 and the GABA transporters40. A recent structural study of a chloride-dependent E290S mutant of LeuT also identified a chloride ion at this location41. Chloride is coordinated in a tetrahedral fashion through residues in TM6 (Ser 320, Gln 316), TM7 (Ser 356) and TM2 (Tyr 69) (Fig. 3a). Interestingly, the hydroxyl group of Ser 320 bridges the Na1 and Cl− sites and is positioned to interact with both ions. The mean ion–ligand distances at the Cl− site are 3.0 Å (Supplementary Table 3) and the B-factors of surrounding atoms are similar to that of chloride, supporting the placement of chloride at this site.

The sodium at the Na2 site is located ‘below’ the plane of the drug towards the cytoplasmic face, in between TMs 1 and 8, and is coordinated in a trigonal bipyramidal fashion by main-chain carbonyls from Gly 42 (TM1a), Val 45 (TM1-hinge) and Leu 417 (TM8), and the side-chain oxygens from Ser 421 and Asp 420 (TM8) (Fig. 3b). The mean ion–oxygen distances are 2.4 Å, in line with reported values for sodium coordination in solution (Supplementary Table 3). Although the interconnected network of interactions between TMs 1, 6, nortriptyline, sodium and chloride provides a structure-based mechanism for the coupling of ion and inhibitor binding42, we do not have a comprehensive understanding of the ion dependence of inhibitor binding in NSSs.

Cholesterol-binding site

A cholesterol molecule is lodged in a trough-shaped cavity bordered by TM5, TM7 and TM1a at a depth equivalent to the inner leaflet of the membrane (Fig. 4a). Branched aliphatic residues are primarily involved in forming the protein–cholesterol interface (359 Å2), thus allowing cholesterol to bury ∼57% of its solvent-accessible surface area. Fo – Fc density for this site clearly demarcated the orientation of the isooctyl group of cholesterol anchored at the junction of TMs 5 and 7 by residues Leu 276, Leu 277 and Ile 358. The β-face of the sterol ring primarily faces residues Tyr 273, Leu 270 and Trp 266 in TM5 and also interacts with residues Val 34, Leu 37, Leu 38 and Ile 41 on TM1a. The α-face of cholesterol interfaces with residues Leu 347 and Ile 351 in TM7 (Fig. 4a).

Figure 4: Cholesterol site.
figure 4

a, Cholesterol (yellow sticks) shown with Fo – Fc density (light magenta) contoured at 2.0σ. Residues that interface with the cholesterol group are represented as sticks. b, Potential role of cholesterol in maintaining an outward-open state of transporter. Cholesterol (sticks with transparent surface) sterically clashes with the position of TM1a in the inward-open conformation of LeuT (PDB code 3TT3)17.

PowerPoint slide

Full size image

Cholesterol has an important role in modulating the function of NSS members43,44, stabilizing an outward-open state of DAT with a concomitant increase in maximum binding or Bmax for cocaine45. In LeuT, TM1a undergoes a large conformational change upon transition from the outward-facing open and occluded states to the inward-open state17. If a similar conformational change were to occur in DAT, it would entirely disrupt the cholesterol site (Fig. 4b). We propose that one mechanism for the action of cholesterol on DAT is that by occupying its binding site in the outward-open, inhibitor-bound state, cholesterol stabilizes the outward-open conformation of the transporter45.

Extracellular and cytoplasmic gates

The ion and ligand binding sites in DATcryst are accessible to solvent from the extracellular face owing to the open gate above the primary binding pocket. The distance between Tyr 124 of TM3 and Phe 319 in TM6a is 10 Å, whereas in the substrate-bound, occluded state of LeuT the corresponding distance is half as long (Fig. 5a, b). Similarly, the 10 Å separation between Arg 52 on TM1b and Asp 475 on TM10 renders the primary binding site accessible to extracellular solution. The steric bulk of the tricyclic moiety combined with the extended N-methylpropylamine chain of nortriptyline prevents both TM1b and TM6a from approaching TM3 and TM8 to cap the putative substrate pocket and close the gate.

Figure 5: Extracellular and cytoplasmic gates and the C-terminal latch.
figure 5

a, Relative locations of the open extracellular gate (red box), closed cytoplasmic gate (blue box) and C-terminal latch (green box) in DATcryst. b, The width of the extracellular gate is depicted by the distances between Tyr 124 and Phe 319 (10 Å), and Arg 52 and Asp 475 (10 Å). Nortriptyline, ions and helices are coloured as in Fig. 1. c, The cytoplasmic gate is closed by polar and electrostatic interactions between TM1a, TM2, IL1, TM6b and TM8. d, The C-terminal helix following TM12 is bound to the cytoplasmic face of the transporter via polar interactions with IL1. Polar and electrostatic bonds are represented as grey dashed lines.

PowerPoint slide

Full size image

In contrast to the extracellular gate, extensive polar interactions at the intracellular face of the transporter form a thick barrier of ∼24 Å between the ligand and ion pockets and solvent to keep the cytoplasmic gate shut. At the cytoplasmic face of the transporter, the indole nitrogen of Trp 30 caps the carbonyl oxygen of Tyr 331 in TM6b, and Arg 27 forms a salt bridge to Asp 435 of TM8 (Fig. 5c). Arg 27, Trp 30 and Asp 435 are strictly conserved in NSS orthologues and LeuT, suggesting that these intracellular gate interactions are general and important facets of the transport mechanism for this family of sodium symporters46,47. Tyr 334, the residue corresponding to Tyr 335 in human DAT, was previously shown to be responsible for shifting the conformational equilibrium of the DAT towards an inward-open state48.

C-terminal latch

Two novel attributes at the C terminus of DATcryst were immediately evident from the structure. Helix 12 is shifted by 22° in comparison to its position in LeuT, resulting in the exposure of TM3 to solvent and lipid (Supplementary Fig. 5). Pro 572, conserved in most eukaryotic NSS members, is probably at the root of the kink between TMs 12a and 12b, and thus has an important role orienting the second half of the helix away from the rest of the transporter. The hairpin between TM12b and the intracellular C terminus of DATcryst is stabilized by hydrogen bonding between the ε-nitrogen of Arg 589 and carbonyl oxygen of Thr 582 (Fig. 5d). The second feature is the C-terminal helix, which contains 2.5 turns from residues 586 to 595, where several hydrogen bonds and a cation-π interaction between Trp 597 and Arg 101 restrain this C-terminal helix near intracellular loop 1 (IL1) at the cytoplasmic face of DATcryst. Although sequence conservation within TM12 and the C terminus is rather low across NSS orthologues, Gly 584 is located at the hairpin hinge and strictly conserved, and only Lys or Arg is present at the position equivalent to Arg 589 of DATcryst, suggesting that the conformation of the C terminus in the structure is a conserved feature among NSS orthologues. Studies of human DAT have identified the region following TM12 to contain sites for protein kinase C-mediated endocytic trafficking3,49, and it is plausible that phosphorylation may alter the conformation or accessibility of the C terminus, allowing it to interact with cellular machinery for internalization. We also note that the latch participates in interactions with IL1, which in turn interacts with TM1a, thus suggesting that the C-terminal latch may modulate transporter activity.

Conclusion

The structure of DATcryst captures the transporter in an inhibitor-bound, outward-open conformation. The TCA nortriptyline targets the primary substrate site and stabilizes the open conformation by sterically preventing closure of the extracellular gate (Fig. 6a). One chloride and two sodium ions are located adjacent to the ligand, suggesting that the binding of ions and inhibitor are directly coupled. A cholesterol molecule bound to a crevice flanking TM1a probably stabilizes the outward-open, inhibitor-bound conformation (Fig. 6b). The structure reveals a C-terminal latch that makes extensive interactions with the cytoplasmic face of the transporter, proximal to the cytoplasmic gate, and thus in a position to modulate transport activity. Taken together, the structure of a eukaryotic DAT reveals novel insights into antidepressant recognition and structural elements implicated in the regulation of neurotransmitter transport, providing a foundation for drug design strategies.

Figure 6: Mechanisms of antidepressants and cholesterol.
figure 6

a, The TCA nortriptyline (magenta) wedges between scaffold helices 3 and 8 and the core helices 1 and 6, preventing the movement of TMs 1b and 6a from closing the extracellular vestibule. b, Cholesterol (yellow) is bound in an intracellular pocket and hinders the movement of TM1a, thereby stabilizing the outward-open conformation of DAT. The C-terminal latch interacts with IL1 as part of the cytoplasmic gate.

PowerPoint slide

Full size image

Methods Summary

The Drosophila DATcryst construct (Supplementary Fig. 1) was expressed in virus-infected mammalian cells and purified by affinity and size-exclusion chromatography. Fab 9D5 was added before crystallization along with nortriptyline (1 mM) at a DAT:Fab molar ratio of 1:1.1 and concentrated down to 3 mg ml−1. Crystals of the complex were obtained in the presence of 100 mM glycine, pH 9, and 38% polyethylene glycol (PEG) 350 monomethyl ether (MME). The structure was solved by molecular replacement using a polyalanine model of LeuT (PDB code 3F3A) and an ensemble of Fab variable and constant domains. Data processing, model building and refinement were performed using standard crystallographic software (Supplementary Table 1).

Online Methods

Screening, construct optimization and protein expression

The Drosophila DAT was selected as a promising candidate for structural studies after screening multiple orthologues of DATs and NETs by fluorescence-detection size-exclusion chromatography (FSEC)50. In addition, FSEC was used to screen other parameters such as detergent efficacy, thermostability, lipid effects, tertiary epitope-specific monoclonal antibodies and sample homogeneity following purification. The DNA encoding the D. melanogaster DAT was provided by S. Amara and was modified by removal of the first 20 amino acids (Δ1–20), by a deletion in EL2 (Δ164–206) and by point mutations to enhance thermostability (V74A, V275A, V311A, L415A, G538L) by PCR-based methods. This modified DAT sequence, deemed DATcryst, was fused to a C-terminal GFP–His8 tag with a thrombin cleavage site (LVPRGS) in place of residues 602–607. DATcryst–GFP–His8 was produced by virus-mediated expression in mammalian cells51,52,53.

Antibody production

Monoclonal antibodies against DATcryst were raised by D. Cawley using standard methods. Antibodies were screened by FSEC and western blot to select clones that recognized natively folded DATcryst protein. Sequencing of Fab regions was performed on mouse hybridoma cells (Fusion Antibodies) and on the intact antibody protein by Edman degradation (by M. A. Gawinowicz). Antibody was purified from hybridoma supernatant using 4-mercapto-ethyl-pyridine resin. Fab protein was generated by papain cleavage of full-length antibody, followed by Fc capture on Protein A resin and cation exchange. Fab was stored in 20 mM sodium acetate, pH 5, 250 mM NaCl and 10% glycerol.

Purification of DATcryst

Membranes were solubilized in TBS (20 mM Tris, pH 8, 150 mM NaCl) containing n-dodecyl-β-d-maltoside (DDM) at a w/w ratio of 0.1 g detergent per 1 g membrane. The detergent-soluble fraction was incubated with cobalt-charged metal ion affinity resin, and DATcryst–GFP–His8 was eluted with 100 mM imidazole in 20 mM Tris, pH 8, 300 mM NaCl, 5% glycerol, 14 μM lipids (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-(1'-rac-glycerol) (POPG) at a weight ratio of 3:1:1), 1 mM DDM and 0.1 mM cholesteryl hemisuccinate . After thrombin digestion to remove the GFP–His8 tag, DATcryst was isolated by size-exclusion chromatography in 20 mM Tris, pH 8, 100 mM NaCl, 5% glycerol, 14 μM POPE, 4 mM decyl-maltoside and 0.1 mM cholesteryl hemisuccinate. The purified DATcryst protein was mixed with Fab 9D5 at a molar ratio of 1:1.1 and used for crystallization trials at 3 mg ml−1in the presence of 1 mM nortriptyline.

Crystallization

Crystals grew in 100 mM glycine, pH 9 and 38% PEG 350 MME using a drop ratio of 1 μl protein and 0.5 μl reservoir solution by hanging drop vapour diffusion. Initial crystals appeared at 4 °C after 2 days, reaching full size after 7 days. Crystals were flash frozen in liquid N2 directly and used for X-ray diffraction data collection.

Structure determination

X-ray data were collected at the Advanced Photon Source (Argonne National Laboratory, beamline 24-ID-C). Data were indexed, integrated, and scaled using HKL2000 (ref. 54) (Supplementary Table 1). The structure was solved using molecular replacement, with ensembles of constant (constant domains of heavy and light chains as one set) and variable domains (variable domains of heavy and light chains as a second set) of Fab coordinates in the PDB along with a polyalanine model of LeuT (PDB code 3F3A). A multi-model search was done using Phaser55. Initial phases were improved by iterative steps of manual model building, refinement and maximum-likelihood density modification using Coot56, Phenix Refine57 and Phenix Phase and Build58, respectively. Multiple rounds of refinement led to the placement of a majority of main-chain and side-chain atoms for both the Fab and DATcryst. The structure was refined to acceptable R-factors (Supplementary Table 1) with residues 20–24 and 600–605 in DATcryst and 135–138 in the heavy chain unmodelled owing to poor density. Nortriptyline, ions and cholesterol molecules were placed into Fo – Fc density contoured at 2σ or greater in the putative substrate pocket, ion sites and at the periphery of the transporter. Stereochemistry was evaluated using MolProbity59.

Ligand binding and uptake measurements

Scintillation-proximity assays using transporter solubilized in detergent60 were carried out using copper yttrium silicate (Cu-YSi) beads (Perkin Elmer) at 0.5 mg ml−1, 30 nM 3H-labelled nisoxetine (1:9 3H:1H) and 10 nM DATcryst–GFP–His8 protein in the same buffer as that used for size-exclusion chromatography. Unlabelled nortriptyline was used as the competitor ligand. Assay plates were read using a MicroBeta TriLux 1450 LSC & Luminescence counter. Data were fitted using a standard single site competition equation, and Ki values were calculated from the IC50 values using the Cheng-Prusoff equation.

Uptake assays were performed using HEK293 cells expressing respective mutant constructs. Cells were re-suspended in 10 μM 3H-dopamine (1:49 3H:1H) containing uptake buffer made with 25 mM HEPES-Tris, pH 7.1, 130 mM NaCl, 1 mM MgSO4, 5 mM KCl, 1 mM CaCl2, 5 mM d-glucose and 1 mM l-ascorbic acid61. Control samples were pre-incubated with 10 μM cold desipramine before addition of label. Assays were quenched with cold uptake buffer containing 1 μM desipramine after 10 min, cells were washed twice with cold uptake buffer and activity was measured from solubilized cells by scintillation counting. Data were plotted using Origin 7.0.

Thermostability screening of DAT

Sites for mutagenesis were selected on the basis of a model of Drosophila DAT built on the template of LeuT and residues were altered to Ala, Leu or Phe21. Individual mutants along with the wild-type construct were transfected into HEK293 cells and kept in culture for 48 h, then tested for binding activity after detergent solubilization. Samples were split and one part was kept at 4 °C, and the other portion of lysate heated at 40 °C for 10 min. 3H-nisoxetine was added before heating to select for mutants that stabilize an inhibitor-bound state of the transporter. Scintillation proximity assay was used to monitor activity in a high-throughput format. Mutants that consistently had an increased melting temperature (Tm) compared to wild-type (Tm = 35 °C) were chosen and pooled into one construct, which yielded a five-mutant construct with a Tm of ∼60 °C.