Main

Signals by the biogenic amine neurotransmitters—dopamine (DA)1, serotonin2 and noradrenaline3—at chemical synapses are terminated by the cognate neurotransmitter sodium symporters (NSSs)4,5,6,7. Biogenic amines play profound roles in the development and function of the nervous system, as well as in animal behaviour and activity; thus NSSs are central to normal neurophysiology and are the targets of a spectrum of therapeutic and illicit agents, from antidepressants and antianxiety medications to cocaine and amphetamines8. Experimental and computational studies have shown that the DA, serotonin (SERT) and noradrenaline (NET) transporters harbour a conserved structural fold9,10, first seen in the structure of LeuT11. Owing to variations in amino acid sequences12, however, the biogenic amine transporters possess distinct yet overlapping pharmacological ‘fingerprints’13.

The dopamine transporter (DAT)14 removes DA from synaptic and perisynaptic spaces, thus extinguishing its action at G-protein coupled DA receptors. To drive the vectorial ‘uphill’ movement of extracellular DA into presynaptic cells, DAT couples substrate transport to pre-existing sodium and chloride transmembrane gradients. Congruent with the multifaceted roles of DA in the nervous system, perturbation of dopaminergic signalling by disruption of native DAT function has profound consequences15,16,17. On the one hand, the amphetamines, potent and widely abused psychostimulants, are DAT substrates that enhance synaptic levels of DA both by competing with DA transport by DAT and by inducing the release of DA from synaptic vesicles into the cytoplasm, from where DA is then effluxed through DAT into the synaptic space18,19,20,21,22,23,24. On the other hand, the Erythroxylum coca leaf-derived alkaloid, cocaine, as well as synthetic cocaine derivatives are competitive inhibitors of DAT and enhance extracellular DA concentrations by locking the transporter in a transport inactive conformation14,25,26,27. Widely prescribed antidepressants specifically inhibit serotonin and noradrenaline uptake and typically have weaker affinities towards DAT28,29.

Mutagenesis, chemical modification, binding and transport studies have implicated the central or S1 binding site in DAT, akin to the leucine and tryptophan site in LeuT, as the binding site occupied by DA, amphetamines, cocaine and antidepressants25,26,30. Moreover, the X-ray structure of a transport-inactive Drosophila melanogaster DAT (dDAT) in complex with nortriptyline shows the antidepressant bound at the central site9,31. Nevertheless, none of these studies have visualized the binding of DA, amphetamine or cocaine to an active DAT, nor have they illuminated distinctions in ligand pose and transporter conformation between substrates and inhibitors. Here we present X-ray structures of dDAT with substrates DA, methamphetamine or d-amphetamine, with the DA analogue 3,4-dichlorophenethylamine (DCP), and with cocaine or cocaine analogues.

Resurrection of transport activity

The previously reported structure of the dDAT-nortriptyline complex exploited a transport-inactive variant with five thermostabilizing mutations (dDATcryst)9. We recovered transport function yet retained favourable crystallization properties by reverting three thermostabilizing mutations (V275A, V311A and G538L) to their wild-type identities and by shifting the deletion of extracellular loop 2 (EL2; Extended Data Fig. 1). This minimal functional construct, dDATmfc, has a melting temperature of 48 °C32, exhibits DA transport with a KM of 8.2 ± 2.3 μM and Vmax of 2.4 ± 0.2 pmol min−1 per 106 cells, compared to wild-type dDAT (dDATwt) with a KM of 2.1 ± 0.7 μM and Vmax of 4.5 ± 0.4 pmol min−1 per 106 cells (s.e.m., Fig. 1a). The dDATmfc construct binds nisoxetine with a Kd of 36 nM compared to a Ki of 5.6 nM for wild-type dDAT31 (Extended Data Fig. 2).

Figure 1: Dopamine occupies central binding site.
figure 1

a, Michaelis–Menten plots of specific DA uptake by dDATwt (squares) and dDATmfc (triangles). Graph depicts one representative trial of two independent experiments. Error bars represent s.d. values for technical replicates measured in triplicate. b, Surface representation of the dDATmfc–DA complex viewed parallel to the membrane, with DA displayed as cyan spheres. Residues Y124 and F319 are shown as cyan sticks on the left and right sides of DA, respectively. c, Chemical structure and Fo − Fc density for DA (3.0σ). d, Close-up view of DA in the binding pocket with hydrogen bonds shown as dashed lines. Sodium ions and water are shown as purple and red spheres, respectively.

PowerPoint slide

Full size image

The central binding site in DAT, NET and SERT can be divided to subsites A, B and C29,33. Subsites A and C are well conserved in dDAT versus human DAT (hDAT), whereas subsite B, a pocket sculpted by TMs (transmembrane helices) 3 and 8, differs from hDAT in that residues lining this pocket in dDAT are Asp121 and Ser426 (Extended Data Fig. 3). We introduced mutations D121G (TM3) and S426M (TM8) into the dDATcryst and dDATmfc constructs to mimic hDAT subsite B33. These mutations enhanced the affinities for nisoxetine, β-CFT (2β-carbomethoxy-3β-(4-fluorophenyl)tropane) and DCP (Extended Data Figs 2, 4). Although constructs harbouring subsite B substitutions improved crystallization propensity, transport activity was extinguished (Extended Data Fig. 3c). Nevertheless, structures bearing these mutations were solved in complexes with cocaine, β-CFT, RTI-55(2β-carbomethoxy-3β-(4-iodophenyl)tropane) or DCP (Supplementary Table 1). In the cocaine, RTI-55 and DCP complexes, superposition of structures with subsite B mutations onto structures of dDATmfc complexes did not reveal prominent structural changes in the binding pocket or deviations in the positions of bound ligand (Extended Data Table 1).

Dopamine bound to central binding site

The structure of dDATmfc bound to DA displays an outward-open conformation (Fig. 1b) where DA is situated in the central binding site, surrounded by TMs 1, 3, 6 and 8. The amine group points towards subsite A and interacts with the carboxylate of Asp46 at a distance of 3 Å. The catechol group occupies a subsite B cavity sculpted by Ala117, Val120, Asp121, Tyr124, Ser422 and Phe325 (Fig. 1c, d), residues predicted to interact with DA using homology models of DAT based on the substrate-bound occluded state of LeuT, analysis of uptake kinetics, and cysteine labelling studies10,25,26,34. Despite a lack of steric interference from DA, Phe319, which is equivalent to Phe253 in LeuT (in which it occludes solvent access to the LeuT binding pocket), remains splayed away from DA, in an orientation seen in the nortriptyline-bound structure9,11.

The location and interactions of DA with dDAT recall predictions made through homology models of hDAT, although discrepancies between the cocrystal structure and the homology models are also evident. Site-directed mutagenesis and molecular dynamics simulations pointed to the crucial role of Asp46 (Asp79 in hDAT) in the recognition of DA10,25,34,35, a residue conserved amongst biogenic amine neurotransmitters. GAT (γ-amino butyric acid transporter), GlyT (glycine transporter), and LeuT contain glycine at the equivalent position owing to the presence of a compensatory carboxylate group in the substrates11. One notable difference between the dDATmfc–DA structure and previous hDAT models, however, is a rotation in the χ1 torsion angle of ∼130° in the side chain of Asp46 to maintain a 3 Å distance to the amine group of DA. This rotation severs the indirect coordination between Asp46 and the sodium ion at site 1 as observed in the nortriptyline–dDATcryst complex9. Unlike previous simulation studies where DA binds in a dehydrated pocket34, we observed non-protein electron density 3.1 Å from the amine group of DA, into which a water molecule was modelled. This water molecule forms a hydrogen bond with the water molecule that coordinates the sodium ion at site 1, resulting in a molecular network that links DA to the ion-binding sites. In the case of LeuT, substrate interaction with sodium site 1 is direct, with the carboxylate of leucine coordinating the sodium ion11.

The catechol ring of DA interacts with TMs 3 and 8 by hydrogen bonds with the carboxylate group of Asp121 (Fig. 1d). Modelling studies predicted that DA occupies multiple poses within the binding pocket with the meta-hydroxyl either interacting with residues in TM8 equivalent to Ser421 or to Ser422 (dDAT numbering)34. By contrast, in the structure of dDATmfc–DA, the para-hydroxyl group interacts with both the carbonyl oxygen of Ala117 and the carboxylate of Asp121 at distances of 2.8 and 3.1 Å, respectively, whereas the meta-hydroxyl group interacts with the side chain of Asp121 at a distance of 2.7 Å and faces Ser422 in TM8 at a distance of 3.8 Å.

The residues poised to interact with the catechol ring vary across DAT orthologues with invertebrate DAT orthologues retaining alanine and aspartate at positions equivalent to residues 117 and 121, respectively, whereas in most mammalian DATs residue Asp121 is replaced by glycine and Ala117 by serine9, the latter of which could act as a surrogate hydrogen bond partner for the catechol group. With hNET, the equivalent residues at 117 and 121 are Ala and Gly, raising the possibility of the catechol group of noradrenaline interacting with the hydroxyl of Ser420 in hNET (TM8; equivalent to Ser422 of dDAT and Ala423 in hDAT). We propose that these compensatory variations within subsite B dictate catecholamine recognition common to both DAT and NET. To explain why NET binds noradrenaline and DA with nearly equal apparent affinity yet DAT prefers DA31, we must invoke longer-range, indirect interactions, perhaps involving subsite B and the non-helical TM6a–6b ‘linker’ because the ‘Phe-box’ surrounding the β-carbon position is conserved between NET and DAT.

Recognition of d-amphetamine and methamphetamine

To understand how amphetamines are transported by DAT despite lacking the hydroxyl groups of the catecholamines, we characterized the interactions between amphetamines and dDAT by binding assays and crystallographic studies. (+)-methamphetamine displaces [3H]nisoxetine binding to dDATmfc with a Ki value of 31 μM, whereas d-amphetamine, which lacks the N-methyl group present in methamphetamine, has a Ki of 86 μM (Fig. 2a). d-amphetamine is 10 to 100-fold weaker in its ability to inhibit DA transport in the dDAT compared to its mammalian counterparts31. The weaker affinities of dDAT for amphetamines compared to mammalian DATs may be due in part to differences in residues of subsite B. Indeed, the presence of Asp121 and Ser426 in invertebrate DATs creates a polar environment that does not complement the non polar benzyl groups of amphetamines (Fig. 1d). In mammalian DATs both methamphetamine (Ki = 0.5 μM) and d-amphetamine (Ki = 0.6 μM) are nearly as effective as cocaine (0.2 μM) at inhibition of DA uptake36. Although the maximal rate of transport (Vmax) for d-amphetamine in hDAT is fivefold lower than for DA, the KM values range from 0.8 to 2 μM22, consistent with the notion that mammalian subsite B is more complementary towards the binding of amphetamines than subsite B in invertebrate DATs.

Figure 2: Amphetamines bind to central site.
figure 2

a, Displacement of bound [3H]nisoxetine by methamphetamine (Ki = 31 μM; squares) and d-amphetamine (Ki = 86 μM; triangles). Graph depicts one representative trial of two independent experiments. Error bars represent s.e.m. values for technical replicates measured in triplicate. b, Surface representation of d-amphetamine–dDATmfc complex viewed parallel to the membrane with ligands shown as a light orange sphere. Residues Y124 and F319 are shown as cyan sticks on the left and right sides of DA, respectively. c, d, Superposition of binding pockets of the d-amphetamine–dDATmfc structure in pale orange with binding pockets of methamphetamine–dDATmfc (c, drug in orange, backbone in grey) and DA–dDATmfc in teal (d). Hydrogen bond interactions are represented as dashed lines. Asp46 undergoes a χ1 torsion angle shift from −168° in DA-bound state to +62° in the d-amphetamine–dDATmfc.

PowerPoint slide

Full size image

The structures of (+)-methamphetamine–dDATmfc and d-amphetamine–dDATmfc displayed outward open conformations with electron densities for the drugs found in the central binding site (Extended Data Fig. 5; Fig. 2b). The amine groups of methamphetamine and d-amphetamine lie closer to Asp46 at subsite A with hydrogen bonding distances of 2.9 Å, and the main chain carbonyl of Phe319 is positioned nearby at 3.3 Å (Fig. 2c). The amine group of d-amphetamine interacts with Asp46, which does not undergo the rotameric shift as seen in the DA-bound structure because d-amphetamine is situated in the centre of the pocket and not displaced by 2.8 Å towards TMs 3 and 8 as seen with DA (Fig. 2d). By contrast with earlier findings25, we do not observe a disruption of the hydrogen bond between Asp46 and Tyr124 despite Asp46 clearly forming a hydrogen bond with the primary amine of d-amphetamine. Phe325 retains edge-to-face aromatic interactions with the phenyl group of amphetamines by way of a contraction of the TM6a–6b linker in comparison to the DA-bound state. Amphetamines adopt poses in the central binding that allow for interactions between their amino and aromatic groups and transporter subsites, thus explaining how the sterically smaller amphetamines compete with DA and act as substrates despite the absence of catechol-like hydroxyl groups.

Dopamine analogue stabilizes partially occluded state

DA is prone to oxidation and thus we sought a stable analogue for crystallographic and biochemical studies. Multiple high-affinity biogenic amine transporter inhibitors harbour halogen groups on the aromatic rings predicted to occupy subsite B33,37 and thus we screened halogenated phenethylamine derivatives for binding to dDAT (Extended Data Fig. 4d). We discovered that 3,4-dichlorophenethylamine (DCP) possessed the greatest affinity and, because it is approximately isosteric to DA, was selected for further study (Fig. 3a).

Figure 3: DCP induces partial occlusion of central site.
figure 3

a, DCP and DA inhibit binding of [3H]nisoxetine to dDATmfc with Ki constants of 4.5 and 8.3 μM, respectively. Graph depicts one representative trial of two independent experiments, and data points indicate average values of technical replicates measured in triplicate for DA (squares) and DCP (triangles). Error bars represent s.d. values for individual data points measured in triplicate. b, Surface representation of the DCP–dDATmfc complex viewed parallel to the membrane with DCP shown as blue spheres and chloride in green. Residues Y124 and F319 are shown as cyan sticks on the left and right sides of DCP, respectively. c, Close-up view of DCP in the binding pocket showing the hydrogen bond between the primary amine of DCP and D46 with a distance of 3.2 Å. Sodium ions are shown as purple spheres. d, Superposition of DA- and DCP-bound dDATmfc structures reveals a 3.1 Å displacement in ligand position. The DA–dDATmfc and DCP–dDATmfc structures are coloured grey and orange, respectively. e, Conformational changes of phenylalanine side chain positions in the partially occluded state of DCP–dDATmfc compared to the outward open state of nortriptyline–dDATcryst, coloured orange and grey, respectively.

PowerPoint slide

Full size image

One DCP molecule is lodged in the central binding pocket, with the amine group forming a hydrogen bond with Asp46, and the dichlorophenyl ring bordered by Val120 and Phe325 in subsites B and C, respectively (Fig. 3b–d, Extended Data Fig. 6). The position of DCP in the pocket is closer to that of d-amphetamine in dDAT and of substrate leucine in LeuT11, whereas the position of DA is shifted towards TMs 3 and 8, probably owing to hydrogen bonding between the catechol hydroxyl groups and Asp121. As a result, the position of Asp46 in the DCP-bound structure is superimposable with the positions seen in the amphetamine-bound structures. Interestingly, the side chain of Phe319 rotates to occlude the binding pocket, a conformational change not seen in any of the inhibitor- or DA-bound structures (Fig. 3c, d). This rotamer of Phe319 prevents solvent access to the pocket, leaving an aperture only ∼1 Å wide on the extracellular side of DCP. The equivalent residue in LeuT, Phe253, adopts the same orientation in the occluded substrate-bound form, supporting the notion that the rearrangements seen in the DCP–dDATmfc structure are on the pathway to a LeuT-like occluded state of dDAT11. Unlike the DA-bound state, there is no evidence of a water molecule associated with DCP in the structures, suggesting that formation of an occluded state is associated with dehydration of the binding pocket (Fig. 3c, d), similar to LeuT.

The partially occluded binding pocket of the DCP–dDATmfc structure primarily results from rotations of TM1b and 6a ‘into’ the binding pocket, towards scaffold helices 3 and 8 and around axes centred near the non helical regions of TMs 1b and 6a (Extended Data Fig. 6a, c). In LeuT, studies of the transporter in solution and in the crystal show that TMs 1b and 6a, along with EL4, undergo conformational changes to close the extracellular gate11,38,39. Comparisons with the outward-open nortriptyline-bound state of dDAT9 and the occluded state of LeuT indicate that, although the DCP ligand is nearly inaccessible to the extracellular solution, the inward rotations of TMs 1b (5.6°) and 6a (∼7°) are less pronounced in dDAT than in LeuT11 (Extended Data Fig. 6f, g). Nevertheless, these helical rotations position the side chain of Phe319 over the extracellular face of the binding pocket, and are associated with a series of phenylalanine side chain reorientations (Fig. 3e). TM11 undergoes an outward movement of 6° to accommodate these side chain shifts, and an inward movement of 5° is observed in TM2 (Extended Data Fig. 6b, d). Interestingly, TMs 2, 7 and 11 at the inner leaflet of the plasma membrane form a second cholesterol binding site (site 2) where a density for cholesteryl hemisuccinate (CHS) was observed in all the reported structures. Indeed, CHS enhances dDAT inhibitor affinity and we speculate that, in native membranes, cholesterol binds to sites 1 and 2, perhaps stabilizing DAT in an outward-open state (Extended Data Fig. 7)39,40,41.

The occluded state is a Michaelis–Menten-like transport intermediate for LeuT and related transporters such as BetP, Mhp1 and MhsT42,43,44 when bound to substrate. Disparities in conformations observed between LeuT and dDAT in the presence of substrate, however, may reflect bona fide differences that are dependent on the relative stabilities of distinct substrate-bound states. Nevertheless, we cannot exclude crystal lattice effects, Fab binding, lipid, or solutes present in the crystallization solutions that may favour the outward-open conformation observed for the dDAT–substrate complexes reported here.

Cocaine binds in the central site

The structure of the dDATmfc–cocaine complex exhibits an outward-open conformation (Fig. 4a) with cocaine bound to the central pocket at a site overlapping the nortriptyline site45, adjacent to the Na1 and Na2 sodium ions and the chloride ion. The tertiary amino group of cocaine forms a salt bridge with Asp46 (TM1b) (Fig. 4b). We observe that the TM6a–6b linker contracts, allowing Phe325 to form edge-to-face aromatic interactions with the benzyl ring of cocaine (Fig. 4c).

Figure 4: Multivalent binding of cocaine.
figure 4

a, Surface representation of the structure of cocaine–dDATmfc viewed parallel to the membrane, with cocaine displayed as yellow spheres. Residues Y124 and F319 are shown as cyan sticks on the left and right sides of cocaine, respectively. b, Close-up view of cocaine in the binding pocket with residues interfacing with cocaine shown as spheres. The tertiary amine of cocaine is 3.4 Å from the carboxylate of D46. c, Superposition of the binding pocket of the d-amphetamine–dDATmfc structure in pale orange with the binding pocket of the cocaine–dDATmfc structure in teal. d, Displacement of [3H]nisoxetine by cocaine for dDATmfc (squares) and dDATmfc subB (triangles) constructs with inhibition constant (Ki) values of 33 ± 3 µM and 3 ± 0.3 µM, respectively. Graph depicts one representative trial of two independent experiments, and data points indicate average values of technical replicates measured in triplicate.

PowerPoint slide

Full size image

The negative electrostatic potential of subsite B in dDAT compared to mammalian DATs probably underlies the reduced affinity of dDAT for cocaine. Altering this charged pocket to mimic mammalian DATs with the mutations D121G and S426M yielded a dDAT construct with enhanced binding affinities for cocaine and β-CFT, but not for RTI-55 (Fig. 4d, Extended Data Fig. 4a, b). We speculate that the carboxylate group of Asp121 does not form favourable interactions with the 4-fluorophenyl and 4-iodophenyl groups of β-CFT and RTI-55, respectively, perhaps accounting for some of the discrepancies in relative binding affinities between hDAT and dDAT.

Ligand docking studies using a homology model of DAT, in combination with biochemical binding assays and ligand-dependent disulphide trapping experiments, have probed the orientation of cocaine in the central binding site25,26,30. These studies predicted that the fluorophenyl moiety of β-CFT forms a hydrogen bond with the side chain equivalent to Asn125 in dDAT. Furthermore, the methyl ester of cocaine was thought to displace the side chain of Tyr124 to abrogate the hydrogen bond between Tyr124 and Asp46 as a mechanism of inhibiting transporter function. The dDATmfc–cocaine structure contradicts the finding that cocaine disrupts the Asp46–Tyr124 interaction and does not place the side chain of Asn125 in a location where it could interact with the fluorophenyl group of β-CFT.

To validate the cocaine-bound structure and conclusively identify residues that interact with tropane-based ligands, structures of dDAT were solved in the presence of the cocaine analogues β-CFT and RTI-55. Anomalous scattering by the iodide of RTI-55 corroborated the location and placement of the aromatic moiety of cocaine proximal to TMs 3 and 8 (Extended Data Fig. 8a). The Fo − Fc ‘omit’ electron density maps for cocaine, β-CFT, and RTI-55 are consistent with the methyl ester group protruding into the base of the extracellular vestibule without disrupting the Asp46–Tyr124 interaction (Extended Data Figs 5, 8b). The position of β-CFT places the fluoro group 6 Å from the amide nitrogen of Asn125, indicating that Asn125 does not directly participate in binding of β-CFT. Superpositions of the cocaine–dDATmfc, β-CFT–dDATcryst, and RTI-55–dDATmfc structures exhibited overall mean root-mean-square deviation (r.m.s.d.) values below 0.7 Å, and residues that make close contacts with the ligands overlap nearly entirely, indicating that the halide-substituted phenyl groups of β-CFT and RTI-55 do not markedly affect the architecture of the binding pocket (Extended Data Table 1).

Residues in the binding pocket that interact with cocaine are shared by β-CFT and RTI-55, with slight deviations in subsite B owing to the presence of halide substituents on these latter two inhibitors. At the distal end of the ligand, the halophenyl rings of β-CFT and RTI-55 or the benzoate of cocaine form van der Waals interactions primarily with Ala117, Val120, Tyr124, Phe325 and Ser422, all of which were previously predicted to interact with these tropane-based inhibitors25,46,47 (Fig. 4b). A comparison with the antidepressant nortriptyline–dDATcryst structure9 reveals that the smaller benzoate group of cocaine is accommodated with a shift of the TM6a–6b linker and a rotation of Phe325 into the binding pocket, which was not previously predicted. The tropane rings of cocaine, β-CFT and RTI-55 are bordered by Phe43, Ala44, Asp46, Ala48, Phe319 and Ser421. The side chain of Phe319 maintains a similar orientation as seen in the dDAT–nortriptyline structure owing to the bulky tropane ring and the methyl ester group present in all three tropane inhibitors. Overall, the docking and photolabelling studies predicted the location and orientation of cocaine and its analogues to overlap with the DA-binding site of DAT25,26,30. Our structures validate these findings yet indicate distinctions that have implications for the mechanism of inhibition by tropane ligands. Taken together, structures of dDAT in complex with tropane ligands suggest that inhibition is achieved by combining a free amine with bulky tropane and aromatic moieties to limit conformational movements of the transporter.

Plasticity of the substrate binding pocket

The inhibitor- and substrate-bound structures of dDAT indicate that the binding pocket of dDAT accommodates ligands of varying sizes largely by adjusting the orientation of the discontinuous region of TM6 and the side chains of Phe319 and Phe325 (Fig. 5; Extended Data Table 2a). Ligand recognition by dDAT is bipartite and requires an amine group that interacts with the carbonyl oxygens of Phe43 and Phe319 or the carboxylate of Asp46 in subsite A, in combination with an aromatic group that is stabilized by van der Waals interactions with residues lining a hydrophobic cleft formed by TMs 3, 6, and 8 of subsite B. The substrates DA, d-amphetamine, and methamphetamine each contain one phenyl ring linked to an ethylamine chain, requiring Phe325 to rotate inward to contract the size of the pocket and maintain edge-to-face aromatic interactions with these ligands (Extended Data Fig. 8c, Fig. 5a, e). In the DCP–dDATmfc structure, Phe319 also rotates inward to cover the ligand, and similar to LeuT, this reorientation of Phe319 is required for the formation of the occluded state (Extended Data Fig. 8d, Fig. 5b, e). In order for Phe319 to cover DCP in the pocket, rotation of Phe319 is accompanied by the inward tilting of TM6a to bring the hinge region of TM6 towards the ligand (Extended Data Fig. 6). Accommodation of the single phenyl rings in tropane inhibitors requires Phe325 to assume a position similar to that seen in the DA-bound structure to provide edge-to-face aromatic interactions (Fig. 5c, f). To enlarge the binding pocket for multiple aromatic rings, the side chains of Phe319 and Phe325 are splayed outward as seen in the antidepressant-bound structures (Fig. 5d, f).

Figure 5: Plasticity confers versatile recognition.
figure 5

a–d, Transverse sections of the binding pocket of dDATmfc are shown as surface representations in the DA- (a), DCP- (b), cocaine- (c) and nortriptyline-bound (d) structures. A water molecule (W) is observed in the vicinity of DA in a. e, f, Schematic representing plasticity within the substrate and drug binding pockets in occluded state of DCP–dDATmfc (broken lines) with DA–dDATmfc (solid line) (e) and cocaine–dDATmfc (broken lines) with nortriptyline–dDATcryst (f, grey, PDB ID 4M48) (solid line). Inset represents schematic to show field of view for a–d and e, f.

PowerPoint slide

Full size image

These structures provide a molecular explanation for the distinction between substrates and inhibitors of biogenic amine transporters. Substrates such as DA and amphetamines contain amine and aromatic functional groups at opposing ends of the molecule that interact with both the extracellular gate of TMs 1b and 6a and the scaffold TMs 3 and 8. In contrast, inhibitors exploit the flexibility of the binding pocket to bind to the outward open transporter with high affinity, acting like wedges to lock the transporter in an outward-open conformation. Tropane ligands achieve inhibition by inserting benzyl or halo-phenyl groups into the cavity of subsite B, with the tropane ring oriented to sterically hinder the movement of the extracellular gate. Antidepressants, like nortriptyline, differ from tropane inhibitors by coupling bulky aromatic moieties with an amine group to block conformational flexibility in the transporter.

Conclusions

Structures of dDAT in complex with both substrates and inhibitors emphasize the role of subsites in the pocket of dDAT in defining ligand specificity, a concept that can be expanded to understand variation in pharmacological profiles between biogenic amine transporters33. The DA-bound structure suggests that interactions between the catechol ring and subsite B, together with hydrogen bonding between the amine group of DA and Asp46, drive closure of the extracellular gates TMs 1b and 6a to form the occluded state. d-amphetamine and methamphetamine are bound in a manner distinct from DA in the pocket, and the absence of hydroxyl groups in amphetamines indicates that hydrophobic interactions must be sufficient for amphetamines to interact with subsite B residues and bridge the scaffold TMs 3 and 8 with the extracellular gating helices. The conformations of the DA, amphetamine and DCP-bound complexes probably represent snapshots following substrate and ion binding but before full closure of the extracellular gate, rather than the occluded conformation seen in LeuT bound to its substrates11,48,49. Questions extending from the substrate-bound structures involve the conformational changes required for DAT to transition to other states of the transport cycle and the roles of subsites A and B in neurotransmitter transport by mammalian DATs. Furthermore, the inhibitor-bound structures of dDAT provide a scaffold for addressing the mechanistic distinction between addictive and non-reinforcing analogues of cocaine25,45,50.

Methods

Constructs

dDAT constructs employed in this study include:

subsite B mutations (subB). Denotes the addition of mutations D121G (TM helix 3) and S426M (TM helix 8) in the binding pocket.

dDATwt. Full-length dDAT without thermostabilizing mutations, fused to a C-terminal GFP tag.

dDATdel. Contains an N-terminal deletion from 1–20 (Δ1–20) and a deletion in extracellular loop 2 (EL2) from 164–191, with thermostabilizing mutations V74A and L415A.

ts2 dDATcryst. Contains thermostabilizing mutations (V74A, L415A), Δ1–20, a deletion in EL2 from Δ164–206, and a thrombin site (LVPRGS) replacing residues 602–607 (Supplementary Table 1). Structure reported of this construct with subB mutations in complex with cocaine and RTI-55.

ts5 dDATcryst. Identical to ts2 dDATcryst except that it contains three additional thermostabilizing mutations (V275A, V311A, G538L) (Supplementary Table 1). Equivalent construct containing subB mutations reported for β-CFT.

dDATmfc. Contains thermostabilizing mutations (V74A, L415A), Δ1–20, a modified deletion in EL2, that is, Δ162–202, and a thrombin site replacing residues 602–607.(Supplementary Table 1). Structures in complex with cocaine, RTI-55, d-amphetamine, (+)-methamphetamine, DA and 3,4-dichlorophenethylamine (DCP) are reported. Structure of dDATmfc containing subB mutations reported in complex with DCP.

dDATmfc 201. Identical to dDATmfc except the deletion in EL2 is from Δ162-201. (Supplementary Table 1). Structure of dDATmfc 201 containing subB mutations reported in complex with DCP.

Expression and purification

The dDAT constructs were expressed as C-terminal green fluorescent protein (GFP)–His8 fusions using baculovirus-mediated transduction of mammalian HEK-293S GnTI− cells51,52. Membranes harvested from cells post-infection were homogenized with 1× TBS (20 mM Tris pH 8.0, 100 mM NaCl) and solubilized with a final concentration of 20 mM n-dodecyl β-d-maltoside (DDM) and 4 mM cholesteryl hemisuccinate (CHS) in 1× TBS. Detergent-solubilized material was incubated with cobalt-charged metal affinity resin and eluted with 1× TBS containing 1 mM DDM and 0.2 mM CHS along with 80 mM imidazole (pH 8.0). The GFP–His8 tag was removed using thrombin digestion followed by concentrating the metal ion affinity purified protein. The thrombin-digested protein was subjected to size exclusion chromatography through a Superdex 200 10/300 column pre-equilibrated with buffer containing 20 mM Tris pH 8.0, 300 mM NaCl, 4 mM decyl β-d-maltoside, 0.2 mM CHS and 0.001% (w/v) 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE). Peak fractions greater than 0.5 mg ml−1 were collected and pooled together. Ascorbic acid (25 mM) was added to the protein solution used to crystallize the DA–dDATmfc complex, to serve as antioxidant. All procedures were carried out at 4 °C.

Fab complexation and crystallization

Antibody fragment (Fab) 9D5 was used to complex with the protein at a molar ratio of 1.2 (Fab):1 (protein). The Fab–DAT complex solution was incubated with 1–2 mg solid drug for 30 min on ice followed by concentration in a 100 kDa cutoff concentrator to 3.5–5 mg ml−1. The concentrated protein was spun down to remove excess drug and insoluble aggregates and plates were set up by hanging drop vapour diffusion. Crystals of Fab–DAT complex grew primarily in conditions containing PEG 400 or PEG 350 monomethyl ether or PEG 600 as the precipitant (Extended Data Table 2b). The pH range of crystallization was 8.0–9.0 for dDATcryst-based constructs and in the range of 6.5–8.0 with dDATmfc-based constructs (Extended Data Table 2b). Crystals of dDATmfc were primarily obtained by streak seeding with a cat whisker dipped in crystals formed with subB containing constructs, 2–5 days after drops were set up. All crystals were grown at 4 °C.

Data collection and structure refinement

Crystals were directly flash-cooled in liquid nitrogen when the PEG 400 concentration in the mother liquor exceeded 36%. For crystals grown in wells containing less than 36% PEG 400, crystals were transferred into cryoprotection solution identical to the reservoir solution but with 40% PEG 400. In conditions with 30–34% PEG 600 as the primary precipitant, 10% of ethylene glycol was added to provide additional cryoprotection. Data were collected either at ALS (5.0.2; 8.2.1) or APS (24-IDC and IDE). Anomalous data for iodine containing RTI-55 complexed with dDAT was collected at 1.6 Å as described in the crystallographic data table. Data were processed using either HKL200052,53 or XDS54. Molecular replacement was carried out for all data sets using coordinates 4M48 with Fab 9D5 and dDATcryst used as independent search models, using PHASER in the PHENIX software suite55,56. Iterative cycles of refinement and manual model building were carried out using PHENIX and COOT57, respectively, until the models converged to acceptable levels of R-factors and stereochemistry.

Radiolabel binding and uptake assays

All binding assays were carried out by scintillation proximity assay (SPA) method58. Reactions contained 5–20 nM protein, 0.5 mg ml−1 Cu-YSi beads, SEC buffer, and [[3H]nisoxetine] from 0.1 to 300 nM for saturation binding assays. Competition binding assays were done with 30 nM [3H]nisoxetine and increasing concentrations of unlabelled competitor. Ki values were estimated from IC50 values using the Cheng–Prusoff equation. Fits were plotted using Graphpad Prism v4.0.

For uptake assays HEK 293S cells were infected with baculovirus expressing dDATmfc or dDATwt, and sodium butyrate was added to 10 mM 8–12 h post-infection. At 24 h, infected cells were adhered to Cytostar-T plates for 2 h, after which the culture media were replaced with uptake buffer (20 mM HEPES pH 7.4, 120 mM NaCl, 5 mM KCl, 2.5 mM CaCl2, 1.2 mM MgSO4, 10 mM d-glucose, 1 mM tropolone, 1 mM l-ascorbic acid59,60. Uptake assays were carried out using [14C]DA over a range of 0.3–60 μM in 100 μl total volume. Samples were read every five minutes for a time course over twenty-five minutes. The linear initial rate of uptake was plotted in the presence and absence of 10 μM nortriptyline to calculate the specific uptake rate. Data were fitted to a standard Michaelis–Menten equation to obtain KM and Vmax values. The significance of specific uptake was assessed at each concentration of DA using a two-tailed Welch’s t-test with 2 degrees of freedom.