Structural insights into GABA transport inhibition using an engineered neurotransmitter transporter

Abstract

γ‐aminobutyric acid (GABA) is the major inhibitory neurotransmitter, and its levels in the synaptic space are controlled by the GABA transporter isoforms (GATs). GATs are structurally related to biogenic amine transporters but display interactions with distinct inhibitors used as anti‐epileptics. In this study, we engineer the binding pocket of Drosophila melanogaster dopamine transporter to resemble GAT1 and determine high‐resolution X‐ray structures of the modified transporter in the substrate‐free state and in complex with GAT1 inhibitors NO711 and SKF89976a that are analogs of tiagabine, a medication prescribed for the treatment of partial seizures. We observe that the primary binding site undergoes substantial shifts in subsite architecture in the modified transporter to accommodate the two GAT1 inhibitors. We also observe that SKF89976a additionally interacts at an allosteric site in the extracellular vestibule, yielding an occluded conformation. Interchanging SKF89976a interacting residue in the extracellular loop 4 between GAT1 and dDAT suggests a role for this motif in the selective control of neurotransmitter uptake. Our findings, therefore, provide vital insights into the organizational principles dictating GAT1 activity and inhibition.

Chimeric inhibitor binding pocket analysis uncovers structural determinants of GABA transporter‐specific pharmacology.

Introduction

The levels of the major inhibitory neurotransmitter, γ‐aminobutyric acid (GABA), are controlled by the activity of GABA transporters (GATs) in the presynaptic and glial cell membranes (Florey & Mc, 1959; Iversen & Neal, 1968). Multiple isoforms of GAT mediate the Na⁺/Cl⁻‐dependent uptake of GABA in neurons, glia, and other cell types (Scimemi, 2014). Exocytic release of GABA in the synapse mediates the generation of inhibitory postsynaptic currents (IPSCs) mediated through the ionotropic GABA_A receptors and metabotropic GABA_B receptors in the postsynaptic neurons (Krnjević & Schwartz, 1967; Bettler et al, 2004; Jacob et al, 2008). GABA is primarily responsible for synchronizing activity among neuronal networks and balances the excitatory neurotransmission mediated by glutamate (Owens & Kriegstein, 2002; Khazipov, 2016). Consequently, impaired GABAergic signaling leads to hypersynchronous excitatory discharges that cause epileptic conditions (Treiman, 2001; Khazipov, 2016). Therefore, control of GABA levels and GABA receptor activation forms a therapeutic basis to alleviate epileptic disorders (Treiman, 2001; Sperk et al, 2004). Besides neuronal roles, GABA is also vital for affecting neural development (Ben‐Ari, 2002), modulation of gastrointestinal function (Hyland & Cryan, 2010), immunomodulation of inflammatory responses (Bhat et al, 2010), and regulation of cardiovascular activity (Bentzen & Grunnet, 2011).

Restoring GABA levels in the neural synapse is possible by inhibiting GABA uptake activity mediated through GAT isoforms (GATs 1, 2, 3 and BGT1; Latka et al, 2020a, Zhou & Danbolt, 2013). Early studies of GABA transport involved the identification of high‐affinity uptake of radiolabeled GABA into slices of rat brain cortex by Iversen and Neal followed by the cloning of rat GAT1 by Kanner et al (Iversen & Neal, 1968; Guastella et al, 1990). GAT1 and GAT3 are the primary neuronal isoforms of the GABA transporter that transports one molecule of GABA in combination with 2Na⁺ and 1Cl⁻ ions resulting in the electrogenic movement of GABA across the synaptic membrane (Kavanaugh et al, 1992; Hilgemann & Lu, 1999; Lu & Hilgemann, 1999). Substrate analogs like β‐alanine, cis‐3‐aminocyclohexanecarboxylicacid (ACHC), nipecotic acid, and guvacine can compete with GABA for uptake (Krogsgaard‐Larsen, 1980; Dalby, 2003). Nipecotic acid and guvacine serve as templates for attaching branched aromatic groups to synthesize competitive inhibitors of GAT activity and facilitate movement across the blood–brain barrier (Ali et al, 1985; Borden et al, 1994; Soudijn & van Wijngaarden, 2000). Among the specific inhibitors of GAT1 activity, tiagabine (Gabitril) is prescribed as medication to treat partial seizures (Suzdak & Jansen, 1995). Analogs of tiagabine, NO711, and SKF89976a are also potent inhibitors of GAT1 mediated uptake and are known to mediate competitive inhibition of GAT activity (Clausen et al, 2006; Zafar & Jabeen, 2018; Latka et al, 2020a).

GATs are structurally related to the biogenic amine and glycine transporters of the solute carrier 6 family (Kristensen et al, 2011). Despite being the first SLC6 member to be cloned and characterized (SLC6A1), the high‐resolution structure of a GAT isoform remains elusive. All members of the solute carrier 6 family are composed of 12 transmembrane (TM) helices that comprise symmetrically organized, discontinuous TMs 1a, 1b and 6a, 6b that gate the substrate and ion transport through alternating access (Joseph et al, 2019). Despite having high sequence and structural similarities with SLC6 members involved in biogenic amine and glycine transport, GATs have distinct substrate recognition propensities and inhibitor binding characteristics. In this context, specific inhibitors of GABA uptake like tiagabine, NO711, and SKF89976a have distinct chemical structures and display no interactions with the biogenic amine transporters that are inhibited with high inhibitory potency by antidepressants and psychostimulants.

Recent structural studies on the Drosophila melanogaster dopamine transporter (dDAT), the human serotonin transporter (hSERT), and the human glycine transporter 1 (hGlyT1) have revealed mechanistic aspects of substrate transport and transport inhibition by inhibitors interacting at the orthosteric and allosteric sites (Penmatsa et al, 2013; Coleman et al, 2016; Shahsavar et al, 2021). In particular, the dDAT has proven to be a versatile model to study interactions of catecholamine neurotransmitters, tricyclic antidepressants, psychostimulants, and chronic pain inhibitors at a high resolution due to the availability of thermostabilizing mutants that render the transporter highly stable and an antibody fragment available for use as a crystallization chaperone (Penmatsa et al, 2013; Wang et al, 2015; Pidathala et al, 2021).

In this study, we attempt to understand the fundamental discrepancies between GABA and catecholamine transporter inhibition through engineering the residues of GAT1 orthosteric site into the primary binding site of a thermostabilized dDAT. In doing so, we observe that the engineered construct of dDAT (dDAT_GAT) interacts with specific inhibitors of GATs including NO711 and SKF89976a, although it lacks GABA transport activity. The high‐resolution X‐ray structures of dDAT_GAT in complex with NO711 and SKF89976a display a reorganized subsite architecture in the primary binding site that reveals the basis for altered inhibitor interaction propensities between GATs and biogenic amine transporters. We further observe an allosteric binding site for SKF89976a in the extracellular vestibule that provides a rationale for its ability to display both competitive and noncompetitive interactions in GATs (Krause & Schwarz, 2005). Using the GAT1 from Rattus norvegicus (GAT1_WT) for biochemical analysis, we validate the structural findings we observe in dDAT_GAT and additionally characterize the importance of the extracellular loop 4 (EL4) in affecting selective substrate entry across SLC6 neurotransmitter transporters.

Results

Engineering GAT1 primary binding site into dDAT

The solute carrier 6 family transporters share a similar architecture but display distinct substrate specificities depending on the nature of the cognate neurotransmitter being transported by the neurotransmitter transporters. Despite high structural similarity, the catecholamine neurotransmitter transporters, dDAT, hDAT, and hNET, share a sequence identity in the range of 41–46% with GAT and GlyT isoforms (Fig EV1). Substrate recognition and competitive inhibitor interactions in neurotransmitter sodium symporters (NSS) occur at the primary binding site where a bulk of the residue differences between the transporters lie. Based on an alignment of the crystallized construct of dDAT with hGAT1 homology model generated using I‐TASSER, we analyzed residues within a 10 Å radius of the primary binding site and facing the binding pocket of dDAT. A similar strategy was successfully employed previously using LeuBAT to understand the interactions of selective serotonin reuptake inhibitors (SSRIs; Wang et al, 2013). The equivalent residues of hGAT1 were substituted in a partly thermostabilized (ts) yet functional construct of dDAT (ts2‐L415A, V74A; Wang et al, 2015). The substitutions include the residues F43Y and D46G around subsite A (Tyr60 and Gly63 in GAT1, respectively), V120L, D121N, S422Q, G425T, and S426V in subsite B (Leu136, Asn137, Gln397, Thr400, and Val401 in GAT1) and A117S, F325L, and V327S in subsite C (Ser133, Leu300, and Ser302 in GAT1; Figs EV2 and 1A). In addition to the primary binding site mutations, an additional mutation was made in the extracellular vestibule in EL4 (E384S). The mutations were carried out progressively and analyzed for homogeneity and expression level using fluorescence‐detection size‐exclusion chromatography (FSEC; Appendix Table S1 and Fig S1A–H). While some construct combinations suffered from poor expression and lack of homogeneity, the final construct carrying 11 mutations (dDAT_GAT) yielded a homogenous FSEC profile (Appendix Table S1 and Fig S1H). Despite a homogenous FSEC profile, the ts2 dDAT_GAT was observed to yield a large fraction of aggregated protein upon large‐scale purification that was not amenable for crystallization trials (Appendix Fig S2A). We, therefore, included three additional stabilizing mutations into dDAT (V275A, V311A, and G538L), carrying GAT1‐like mutations (Penmatsa et al, 2013). The purification of ts5 dDAT_GAT construct yielded improved levels of homogenous protein (Appendix Fig S2B) although, as known from earlier studies with thermostabilized dDAT, the transporter is locked into the outward‐open conformation and does not retain any transport activity (Appendix Fig S2C; Penmatsa et al, 2013).

Figure EV1. Multiple sequence alignment of dDAT_GAT with SLC6 members.

Multiple sequence alignment of SLC6 members, hGAT1, hGlyT1, hSERT, hNET, dDAT, and dDAT_GAT, highlighting the extent of sequence conservation (highlighted in blue) in the transmembrane helices marked as colored cylinders above each stretch of sequence alignment. Orange arrows under the dDAT_GAT sequence highlight the five thermostabilizing mutations including V74A, V275A, V311A, L415A, and G538L. The red boxes highlight the positions substituted in dDAT primary binding site and vestibule to resemble GAT1 in the dDAT_GAT construct.

Figure EV2. Subsite architecture of primary binding pocket in dDAT.

The structure of dDAT in complex with reboxetine (PDB id. 4XNX) in its outward‐open conformation is shown. Inset shows the schematic representation of the trilobed subsite architecture of the primary binding pocket in dDAT overlapped with reboxetine in the center displaying subsite A (red), subsite B (blue), and subsite C (yellow). Residues surrounding primary binding site are represented as sticks.

Figure 1. Engineering a GAT1 binding site into dDAT.

• A
  Sequence alignment between hGAT1, hSERT, hNET, dDAT, and dDAT_GAT in the vicinity of the primary binding site. GAT1‐like substitutions (11 mutations) were carried out in dDAT.
• B
  Chemical structures of GABA and substrate analogs, nipecotic acid and guvacine. Structures of specific inhibitors of GAT1 tiagabine, NO711, and SKF89976a.
• C
  Microscale thermophoresis (MST) with purified ts5 dDAT_GAT construct displaying affinity of 10 μM for NO711.
• D
  dDAT_mfc did not show any interactions with NO711 in the MST experiments.
• E
  SKF89976a interacts with purified ts5 dDAT_GAT construct with an affinity of 34 μM.
• F
  MST experiments display no interactions of purified dDAT_mfc with SKF89976a. The MST profiles shown are one of two independent measurements.
• G
  Michaelis–Menten uptake kinetics of ³H‐GABA by the GAT1_WT used in the study displaying a K _M value of 11.4 μM.
• H
  Inhibition of ³H‐GABA uptake by the substrate analog nipecotic acid displaying a K _i value of 14.4 μM.
• I
  Inhibition of ³H‐GABA uptake through GAT1_WT by tiagabine (K _i  = 725 nM), NO711 (1.07 μM) and SKF89976a (7.3 μM).

Data information: All measurements of inhibition potency were the result of two independent experiments performed in triplicate (n = 6), and all the points were used for calculating the values where the error bars represent s.e.m. Control data are included in Appendix Figs S6 and S7.

Source data are available online for this figure.

The purified dDAT_GAT in detergent micelles displays weak interactions with the GAT1 inhibitors NO711 and SKF89976a with a K _d value of 10 and 34 μM using microscale thermophoresis measurements, respectively (Fig 1B–F). Interactions of substrates like GABA and nipecotic acid with dDAT_GAT did not yield a clear binding signal likely due to weaker affinity of substrates to dDAT_GAT in comparison with GAT1. Since the observed binding affinities for ts5 dDAT_GAT were weaker than the GABA transport inhibition potencies reported for GAT1, we employed the GAT1_WT to observe and validate our structural findings with dDAT_GAT. The GAT1_WT displayed GABA transport activity in HEK293S (GnTI^‐) cells with K _M value of 11 μM and inhibition constant toward nipecotic acid at a value of 14 μM (Fig 1G and H). Similarly, inhibitors of GABA uptake like tiagabine, NO711, and SKF89976a (Fig 1B) had inhibition potencies of 0.72, 1.0, and 7 μM, respectively (Fig 1I and Appendix Table S2). The subsequent structural studies in the manuscript have been carried out using ts5 dDAT_GAT, referred to henceforth as dDAT_GAT, in complex with a heterologously expressed fragment‐antigen binding (Fab 9D5) that is essential for crystallizing the diverse constructs of dDAT. The interactions of the Fab are at the cytosolic face of dDAT and do not interfere with the inhibitor interactions with the transporter (Penmatsa et al, 2013, Pidathala et al, 2021; Appendix Fig S2D).

Substrate‐free dDAT_GAT structure reveals altered subsite distribution

The structure of the substrate‐free dDAT_GAT construct was determined to a resolution of 3.2 Å in outward‐open conformation with clear densities being visible for all the side chains including the substituted residues (Figs 2A and EV3A). The primary binding site and the extracellular vestibule leading into it were filled with weak densities into which water molecules were modeled (Fig EV3B). A comparison of the substrate‐free structures of the dDAT_GAT and dDAT_NET constructs display a Cα rmsd of 0.5 Å indicating relatively unaltered main chain conformation between the two structures. Structural changes in the main chain are primarily confined to the substrate/inhibitor binding site, particularly in the region of the TM6 linker (Fig 2B). Substitution of multiple residues in the binding site causes substantial changes to the side chain H‐bonding network that affects the organization of the inhibitor binding site. For instance, in subsite A, a conserved aspartate in TM1b (Asp46^dDAT) in all biogenic amine transporters (Fig 2C) is substituted to a glycine in GATs (Gly63^GAT1), GlyTs (Gly119^GlyT1) and the bacterial amino acid transporter LeuT (Gly24^LeuT). The glycine substitution accommodates the carboxylate group of the substrates that participates in the coordination of Na⁺ ion at site 1, as observed in LeuT (Yamashita et al, 2005). Subsite B, lined by residues in TM3 and TM8, forms a partially hydrophobic cavity in the case of the DAT, NET, and SERT. The GAT1‐like substitutions in this region, D121N, S422Q, G425T, and S426V occlude this subsite through side chain interactions mediated by a network of H‐bonds. In dDAT_GAT, the interactions are between side chain amide groups of Asn121 and Gln422 (2.7 Å), main chain carbonyl of Gln422 (C=O) with γ‐hydroxyl group of Thr425 (2.5 Å), Asn121 γ‐amide group with γ‐hydroxyl group of Thr425 (3.2 Å) and between the main chain of Asn121 (C=O) and side chain of Asn125 (2.4 Å). These interactions substantially minimize the accessible cavity of subsite B in dDAT_GAT, which is a critical region that aids in enhancing the affinities of inhibitors of catecholamine transporters. Major changes in the organization of subsite C are also observed wherein GAT1‐like substitutions alter the position of the TM6 linker (Fig 2D). In earlier studies, we observed the presence of subsite C interactions in a gap between TM6 linker and TM3 that allows inhibitors to have specific interactions, particularly in the case of the norepinephrine transporter (NET; Pidathala et al, 2021). In dDAT_GAT construct, this cavity is blocked due to the widening of the TM6 linker, which is the site of F325L substitution. In the substrate‐free structure of dDAT (PDB.id 6M0F), the disordered TM6 linker comprising residues Gly322‐Pro323‐Gly324 is held in place at a distance of 3.9 Å between the main chain carbonyl of Gly322 (i^th residue) and main chain amide of Phe325 (i + 3 residue; Cα‐Cα distance 6.0 Å). In the dDAT_GAT, this motif is widened with a corresponding distance between main chain carbonyl of Gly322 and amide (‐NH‐) group of Leu325 being 4.8 Å (Cα‐Cα distance 6.8 Å). This likely occurs in case of dDAT_GAT due to the introduction of H‐bond interactions between the Gly322 carbonyl with the hydroxyl group of F43Y substitution and the position of Leu325 shifting further by 1.1 Å due to a new H‐bond interaction of the Leu325 carbonyl group with the side chain hydroxyl group of Ser117 substitution in dDAT_GAT (Fig 2B). These interactions block access to subsite C and open up an alternate subsite within the widened TM6 linker that we refer to as the subsite C′ (Fig 2D). The altered subsite architecture displayed in the engineered dDAT_GAT substrate‐free structure can, therefore, underlie the discrete inhibitor specificities observed in the GATs in comparison with biogenic amine transporters.

Figure 2. Substrate‐free structure of dDAT_GAT displays an altered subsite organization.

• A
  X‐ray structure of the substrate‐free dDAT_GAT (PDB id. 7WGD) displayed with substitutions made within the primary binding site in TMs 1 (deep red), 3 (orange), 6 (green), and 8 (cyan). An additional mutation (E384S) was carried out in the vestibule with a total of 11 substitutions in dDAT to engineer the dDAT_GAT.
• B
  Panel displays an overlay of the substrate‐free structures of dDAT (PDB id. 6M0F) (gray) and dDAT_GAT (PDB id. 7WGD) (light pink) with a Cα rmsd of 0.50 Å. Altered position of the residues along with the angular shifts and displacement of the main chain are depicted. H‐bond networks resulting out of the substitutions within the binding pocket are displayed as dashed lines.
• C
  Conventional trilobed subsite architecture observed in biogenic amine neurotransmitter transporters.
• D
  Altered subsite architecture displayed with the creation of a new subsite C′ with a substantially minimized subsite B accessible to GAT1 inhibitors resulting from the GAT1‐like substitutions.

Figure EV3. Electron densities of substituted side chains and water molecules in the vestibule.

• A
  Substrate‐free dDAT_GAT structure displaying outward‐open conformation. All the mutated side chains display consistent densities. 2Fo‐Fc maps modeled in the panel at sigma level of 1.0.
• B
  Putative densities for water molecules identified and modeled within the vestibule for substrate‐free dDAT_GAT.

NO711 and SKF89976a interact with the engineered primary binding site

In order to analyze the interactions of GAT1‐specific inhibitors with the engineered dDAT_GAT construct, we crystallized dDAT_GAT in complex with NO711 and SKF89976a. The X‐ray structures of the inhibitor‐bound complexes were determined to a resolution of 2.75 and 2.9 Å, respectively (Appendix Table S3). Both inhibitors are analogs of tiagabine and are known to inhibit GABA uptake specifically in GAT1 isoform, with high affinity (Borden et al, 1994). NO711 is a derivative of guvacine and has a diphenylmethylene group that is linked with a 2‐amino oxyethyl group to the guvacine moiety (Fig 1B). The NO711‐bound complex displays an outward‐open state that is typical of most dDAT structures. The density for NO711 could be clearly identified in the primary binding site with the guvacine moiety interacting close to subsite A (Appendix Fig S3A, and Fig 3A and B). The carboxylate group of guvacine occupies the pocket that is created due to the substitution of D46G and participates in the coordination of the Na⁺ ion (3.1 Å) and also interacts with the hydroxyl group of Tyr124 at a distance of 2.9 Å, which is conserved in GAT1 (Tyr140), through a H‐bond. Tyr140 of GAT1 was also implicated in neurotransmitter recognition and substitution at this position, Y140F, led to impaired recognition of GABA (Bismuth et al, 1997), likely due to the interactions that the hydroxyl group could have with the carboxylate of the substrate. The interaction with the inhibitor shifts the Cα‐Cβ dihedral angle of Tyr124 by 34° in comparison with the substrate‐free dDAT_GAT, where it displays water‐mediated interactions with the side chain of Asp475 (TM10). The linker connecting the guvacine with the diphenyl methylene group is in a bent configuration in the binding pocket and accommodates the two aromatic groups away from subsites B and C (Fig 3A and B). The phenyl group that is above the plane of the inhibitor interacts with edge‐to‐face aromatic interactions with Phe319 that could stabilize the outward‐open state in GAT1 (Fig 3A). Alternately and rather interestingly, the second phenyl group is wedged inside the cavity created by the widening of the TM6 linker in the newly observed subsite C′ (Fig 3A). The phenyl group of NO711 forms a network of aromatic and hydrophobic interactions with Phe318, Phe319, and a displaced Leu325 (F325L equivalent to Leu300^GAT1). The displacement of the Leu325 in TM6 linker is clearly a prerequisite for NO711 to bind since the Phe325 in the substrate‐free and nisoxetine‐bound structures of dDAT (PDB.ids 6M0F, 4XNU) would clearly clash with the phenyl group of NO711 and requires displacement for the inhibitor to interact with subsite C′ in TM6 linker region (Fig 3F). The Leu325 is displaced by about 1.3 Å at its Cα position and undergoes a ~70° rotameric shift in the Cα‐Cβ torsion angle to facilitate inhibitor interactions and avoid clashes with the aryl groups of the inhibitors (Fig 3F).

Figure 3. Interactions of competitive inhibitors NO711 and SKF89976a in the primary binding site.

• A
  Sagittal section of the X‐ray structure of the dDAT_GAT in complex with the drug NO711 displays interactions of the guvacine moiety accommodated within subsite A with the carboxylate group displaying interactions with the Na at site 1. The diaryl group attached to guvacine with a long linker interacts within the TM6 linker to form aromatic π stacking interactions with the F319 and plausible CH‐π interactions L325.
• B
  Fo‐Fc (2.4σ), Polder (6.0σ) and composite omit maps (0.8σ) of the NO711 interacting in the primary binding site of dDAT_GAT.
• C
  X‐ray structure surface cutaway (sagittal section) displaying interactions of SKF89976a in the primary binding site and allosteric site. The molecule interacts with a very similar pose as that of NO711. The F319 is, however, in an occluded state unlike the NO711‐bound dDAT_GAT.
• D
  Fo‐Fc (3.0σ), Polder map (5.0σ) and composite omit map (1.0σ) of SKF89976a bound in the primary binding site.
• E
  Structural overlaps of NO711 and SKF89976a‐bound dDAT_GAT display very similar binding pose of the two drugs; Cα rmsd of 0.37 Å.
• F
  Structural comparison of nisoxetine‐bound dDAT (PDB id. 4XNU) and NO711‐bound dDAT_GAT.
• G
  Structural comparison of cocaine‐bound dDAT (PDB id. 4XP4) and SKF89976a‐bound dDAT_GAT.

The structure of the SKF89976a complexed to dDAT_GAT was resolved to a resolution of 2.9 Å that also revealed clear density for the inhibitor bound in the primary binding site (Fig 3C and D; Appendix Fig S3B). In addition to the inhibitor bound in the binding site, an additional density for SKF89976a was observed in the extracellular vestibule (discussed later). SKF89976a is a nipecotic acid derivative, and the inhibitor pose overlaps nearly exactly with that of NO711 except for linker region of the two drugs where NO711 has an additional atom compared with SKF89976a (Fig 3E). The nipecotic acid moiety closely interacts in the vicinity of subsite A displaying coordination of Na⁺ bound at site 1 (3.6 Å) and a H‐bond interaction with Tyr124 hydroxyl group (3.4 Å; Fig 3C). The diphenyl butylene moiety of SKF89976a is surrounded by similar groups in subsite C′ and lacks any interactions in the subsite B, and just like the NO711‐bound structure, the Leu325 is displaced by a torsional angle shift of ~85°. However, unlike the NO711‐bound dDAT_GAT, the position of Phe319 in the SKF89976a‐bound dDAT_GAT is in an occluded state with Phe319 shifting by ~100° in the Cα‐Cβ torsion angle to form an occluded state of the transporter (Fig 3E) that resembles the occluded state of a dichlorophenylethylamine‐bound dDAT (PDB.id 4XPA; Appendix Fig S4). A comparison of the SKF89976a complexed dDAT_GAT structure with cocaine‐bound dDAT (PDB.id 4XP4) reveals clashes of the cocaine's benzoyl ester moiety with the GAT1‐like substitutions around subsite B region with G425T, D121N and S422Q that interact among themselves with H‐bonds. The G425T side chain sterically prevents the interaction of any hydrophobic moiety of the inhibitor at subsite B (Fig 3G). While multiple docking and simulation studies done with inhibitors, interacting with GAT1, display guvacine and nipecotic acid interactions close to subsite A interacting with Na1 and Tyr140^GAT1, the diaryl groups of tiagabine are predicted to dock within the vestibule (Skovstrup et al, 2010; Jurik et al, 2015). The ability of GAT1‐specific inhibitors to interact within the primary binding site with an altered subsite configuration, demonstrated in this study, gives a clear structural glimpse into the inhibitor specificity of GAT1.

Control of substrate and inhibitor interactions at TM6 linker

The structures of inhibitors complexed to dDAT_GAT clearly indicate the importance of the F325L (Leu300 in GAT1_WT) in regulating the interactions of the substrate and inhibitors in the binding pocket. A comparison of this position among SLC6 members revealed that the biogenic amine transporters have a phenylalanine in this position, whereas the GATs have a substitution to a leucine. The GlyTs, which transport an even smaller neurotransmitter, glycine, have a tryptophan substitution at this position (Trp376^hGlyT1 and Trp482^hGlyT2; Fig 4A). In hGlyT2, it was observed that substitution of the Trp482 to leucine or phenylalanine causes reduced substrate specificity to glycine, thereby allowing a wider range of amino acids to interact with the transporter and elicit currents (Carland et al, 2018). The Leu300 in GAT1_WT was, therefore, substituted to multiple residues including a phenylalanine, valine, alanine, and tryptophan to analyze the effects on GABA transport activity. The substitution to L300W in GAT1_WT yielded a nearly inactive transporter (Fig 4B). A comparison of the dDAT_GAT structure with an inhibitor‐bound (cmpd1, a derivative of bitopertin), inward‐open human GlyT1 (PDB.id 6ZBV) and the AlphaFold2 model of GlyT1 in outward‐open state reveals that the Trp376^hGlyT1 would sterically block the interactions of both inhibitors and substrate within the primary binding site (Fig 4A). In the context of the mutations L300V and L300A, the former substitution is also a branched hydrophobic side chain that would allow interactions of GABA within the binding pocket similar to leucine. Further truncation of this aminoacid side chain to L300A would enhance the volume of the binding pocket allowing the substrate GABA greater flexibility in interactions within the binding pocket leading to weakened interaction affinities likely lowering the uptake activity for ³H‐GABA (Fig 4B). We further analyzed the effects of L300F substitution on GAT1 activity and observe that GABA uptake activity is compromised by a twofold increment of K _M to 20.9 μM with a twofold reduction in V _max values (Fig 4C and Appendix Table S2). A similar weakening of nipecotic acid inhibition potency was also observed with a 10‐fold increment in inhibition constant (134 μM) in nipecotic acid competition of GABA uptake (Fig 4D). Contrary to this behavior, the inhibition constant with bulkier inhibitors like tiagabine, NO711, and SKF89976a displayed a higher inhibition potency indicating an enhanced ability to compete with GABA (Fig 4E –G). This is a likely outcome of L300F in GAT1_WT being positioned in a configuration that is similar to the F325L mutant in dDAT_GAT and is displaced enough to allow interaction of the inhibitors with subsite C′ in the environment of the GAT1_WT binding site. This could be a direct consequence of the F43Y and A117S substitutions having H‐bond interactions with the carbonyl groups of Gly322^dDAT and F325L, respectively, in the TM6 linker to facilitate its widening, to allow GAT inhibitor interactions (Fig 2B). This behavior could also be further aided by the reduction of GABA affinity to the L300F that consequently aids in the enhanced ability of the inhibitors to compete for GABA transport.

Figure 4. TM6 linker plays a vital role in GABA uptake.

• A
  Side view of structural overlaps between dDAT_GAT and hGlyT1 in its inward‐open state (PDB id. 6ZBV) with a Cα rmsd of 2.73 Å and hGlyT1 AlphaFold2 model in outward‐open state with a Cα rmsd of 1.43 Å. Insets show the presence of W376 in the position of L300 that occludes the substrate binding within the primary binding site and restrict it to smaller substrates like glycine.
• B
  Substitution of L300 to other hydrophobic side chains and the consequences on transport activity. L300W has the maximal effect on the transport activity. The histogram represents mean of a total six measurements (n = 6) done as two independent measurements. Individual data points are represented as black dots around the error bar that represents s.e.m.
• C–G
  Changes to Michaelis–Menten kinetics for L300F mutant that displays weakened ³H‐GABA uptake activity with K _M  = 20.9 μM. A near 10‐fold loss of nipecotic acid inhibition potency was observed at a value of 134.3 μM. The inhibition potencies of the inhibitors display lowered value suggesting an improved ability to compete for the GABA uptake.

Data information: (C–G) All data plots were performed as two independent experiments each time done in triplicate (n = 6). Error bars in the data display s.e.m. Statistical significance of inhibition potencies was measured between the mutant and GAT1_WT transporter measured with an unpaired t‐test display P values as follows; nipecotic acid, panel E‐0.0001; tiagabine, panel F‐0.0017; NO711, panel G‐0.018; SKF89976a, panel H‐0.0009. GAT1_WT transport kinetics and nipecotic acid inhibition potency are redisplayed from Fig 1G–I for comparison. Control data are represented as Appendix Figs S6 and S7.

Source data are available online for this figure.

Allosteric binding site of SKF89976a

The structure of dDAT_GAT complexed to SKF89976a displayed an additional density at a second site in the extracellular vestibule of the transporter (Fig 5A and B, and Appendix Fig S3C). The density could clearly fit an additional SKF89976a molecule that forced the displacement of Phe319 to an occluded conformation (Fig 5C and D). The aromatic residues in dDAT_GAT Tyr43, Tyr124, and Phe319 form multiple aromatic edge‐to‐face interactions with the phenyl groups of SKF89976a bound at the primary binding site and the allosteric site in the vestibule (Fig 5D). The N‐atom in the nipecotic acid group of SKF89976a interacts with Asp475 directly through a H‐bond (3.0 Å). The residue occurs in TM10 and is also conserved in GAT1 (Asp451) and is implicated in the formation of a thin extracellular gate in the vestibule (Ben‐Yona & Kanner, 2013). The carboxylate group of SKF89976a indirectly interacts through a water molecule with the E384S residue (Ser359^GAT1), which is a substitution we performed in the extracellular vestibule (EL4) of the transporter. A bulk of the residues that surround the SKF89976a in the allosteric site are similar between GAT1_WT and dDAT except for an additional residue insertion (Ser456) in the TM10 of GAT1_WT. The insertion is known to influence efficiency of ion‐coupled transport (Dayan et al, 2017) and is in proximity to the allosteric site and could also influence the interactions of the inhibitor at the allosteric site, in GAT1_WT. A comparison of the SKF89976a in the allosteric site with citalopram‐bound hSERT (PDB.id 5I73) displays broad similarities in the location of the allosteric site in the vestibule. Interestingly, citalopram interacts with hSERT without disrupting the thin extracellular gate between Arg104^hSERT and Glu493^hSERT unlike SKF89976a interactions observed in dDAT_GAT (Fig 5C). We also observe that the residues in the surrounding environment of citalopram‐bound hSERT differ substantially from dDAT_GAT that could dictate the differences in specificity toward the two inhibitors (Fig 5C). We decided to investigate whether the allosteric interaction with SKF89976a can be monitored in GAT1_WT. In order to validate this, we analyzed the alterations in ³H‐GABA uptake kinetics in GAT1_WT with increasing concentrations of tiagabine, NO711, and SKF89976a using Eadie–Hofstee plots (Figs 5E–G and EV4). Tiagabine and NO711 display competitive inhibition with increasing K _M values but unaltered V _max values indicating a clear competitive inhibition propensity with increasing inhibitor concentrations (Figs 5E and F, and EV4). Consequently, increments in SKF89976a displayed both an altered K _M and reduced V _max values leading to a behavior that suggests both competitive and noncompetitive components of inhibition of GAT1 by SKF89976a (Fig 5G). This suggests that SKF89976a does interact at an alternate site in GAT1 leading to allosteric inhibition of GABA uptake. Evidence of this behavior was previously noted in multiple experiments involving SKF89976a interactions in GAT1. Interestingly, SKF89976a was observed to be a useful inhibitor to separate stoichiometric transport current and transmitter‐mediated Na⁺ currents in GAT1 and displayed competitive and noncompetitive modes of inhibiting GAT1 activity (Cammack & Schwartz, 1996; Krause & Schwarz, 2005). It is, therefore, plausible that the SKF89976a binding to the allosteric site in dDAT_GAT construct, observed in this study, could provide a rationale for the abilities of this inhibitor to have altered properties in comparison with competitive inhibitors of GAT1 activity.

Figure 5. Allosteric binding site of SKF89976a.

• A
  Longitudinal cross section of SKF89976a‐bound dDAT_GAT displays proximal interactions of SKF89976a in the allosteric site.
• B
  Fo‐Fc and Polder maps of SKF89976a depict the clear presence of the inhibitor in the vestibule.
• C
  The structures of dDAT_GAT in complex with SKF89976a and hSERT in complex with citalopram are shown. Insets show the interactions of residues present in the secondary binding pocket of dDAT_GAT and hSERT with SKF89976a and citalopram, respectively. SKF89976a displays a different pose in comparison with the citalopram‐bound hSERT but generally coincides with the position of citalopram in the secondary site.
• D
  Relative positions of the two SKF89976a molecules in the allosteric and primary binding sites interact through a network of aromatic and polar interactions.
• E, F
  Eadie–Hofstee plots of (E) tiagabine, and (F) NO711 display competitive mode of inhibition of ³H‐GABA uptake in GAT1_WT.
• G
  Eadie–Hofstee plot of SKF89976a displaying mixed (competitive + noncompetitive) inhibition of ³H‐GABA uptake in GAT1_WT.

Data information: (E–G) Concentrations of the inhibitors were decided based on the K _i values calculated for individual inhibitors. Each data point was plotted using six measurements carried out in two independent measurements in triplicates. Error bars represent s.e.m.

Source data are available online for this figure.

Figure EV4. ³H‐GABA transport kinetics and inhibition by GAT1 inhibitors.

• A–C
  Michaelis–Menten plots of (A) tiagabine and (B) NO711 showing competitive mode of inhibition of ³H‐GABA uptake in GAT1_WT. (C) Michaelis–Menten plots of SKF89976a indicate the mixed mode (competitive + noncompetitive) of inhibition of ³H‐GABA uptake in GAT1_WT. The K _M and V _MAX values with increasing concentrations of the inhibitors are shown in a tabular form. Each data point represents six measurements carried out in two independent experiments in triplicates. The error bars represent s.e.m.

Selective control of neurotransmitter transport at extracellular loop 4 (EL4)

An analysis of the interactions of SKF89976a bound in the allosteric site reveals that the carboxylate in the nipecotic acid group of SKF89976a faces the EL4 residue at E384S (Ser359^GAT1) and indirectly interacts with the hydroxyl of the serine through a weak water‐mediated interaction (3.8 Å; Figs 5D and 6A). The modeled water is also in close proximity (3.2 Å) with guanidine group of Arg52 in TM1b that is conserved in GAT1_WT (Arg69; Fig 5A). The EL4 in the recent past has gained prominence as a lid that can control the association of substrate and inhibitors for selective entry into the extracellular vestibule and interaction of inhibitors in biogenic amine transporters, particularly in the serotonin transporter (Rannversson et al, 2015; Esendir et al, 2021). Novel allosteric modulators of human dopamine transporter (KM322) are also observed to interact in the vicinity of EL4 (Asp385^hDAT), which is the equivalent residue of Glu384^dDAT (Fig 6B; Aggarwal et al, 2019). The proximity of the nipecotic acid moiety of SKF89976a to E384S suggested that the GABA carboxylate and other substrate analogs could interact in a similar fashion (Fig 6A). Superposition of dDAT structure on the dDAT_GAT complexed to SKF89976a revealed that the acidic residue at this position, Glu384^dDAT, could alter the interactions with the substrate and affect its entry into the vestibule (Fig 6A and B). This premise was tested through the substitution of S359E in GAT1 and E384S in dDAT_WT. In GAT1, S359E mutation in EL4a led to a modestly weakened interaction with the substrates indicated by the increased K _M toward GABA, although the V _max of the substituted transporter was higher than the GAT1_WT, which correlated with a proportionate increment in expression level (Figs 6C and EV5A). Interactions of S359E GAT1 with nipecotic acid also displayed a twofold weakening of inhibition potency (30.1 μM) in comparison with the GAT1_WT (14.4 μM; Fig 6D and Appendix Table S2C). On the contrary, the dDAT E384S mutation retained similar expression levels but completely compromised the transport activity, highlighting the importance of EL4 in affecting selective entry of neurotransmitters among catecholamine neurotransmitter transporters but less so in the GABA transporter (Figs 6E and EV5B).

Figure 6. EL4 is involved in neurotransmitter gating.

• A
  Water‐mediated H‐bond interactions displayed in SKF89976a interactions with E384S substitution.
• B
  Alignment shows the presence of acidic side chains in biogenic amine transporters in the EL4.
• C
  Michaelis–Menten kinetics of GAT1_WT and GAT1 S359E displays a weakened K _M value (28.54 ± 2.3 μM). The enhanced V _max is a consequence of the enhanced expression of the mutant transporter in the assayed cells monitored by FSEC.
• D
  ³H‐GABA uptake inhibition by the substrate analog nipecotic acid for GAT1 S359E displaying a K _i value of 30.1 μM.
• E
  The equivalent mutation in dDAT_WT E384S causes a complete ablation of transport activity despite retaining similar expression levels as dDAT_WT.

Data information: (C–E) The kinetic plots were plotted as a mean of six measurements carried out in two independent measurements in triplicate. The error bars represent s.e.m.

Source data are available online for this figure.

Figure EV5. FSEC profiles of EL4 mutants of GAT1 and dDAT.

• A, B
  The FSEC profiles of (A) GAT1 S359E compared with GAT1_WT and (B) dDAT E384S compared with dDAT_WT.

Discussion

The structures of the engineered dDAT_GAT in complex with tiagabine analogs NO711 and SKF89976a reveal the importance of the residue environment within the primary binding site in dictating inhibitor interactions. We observe the structural consequences of GAT1‐like substitutions that shift the trilobed subsite organization for inhibitor interactions to a “bean‐shaped” subsite organization (Fig 7A). This is a consequence of minimized access to subsite B and formation of an altered subsite C′ due to a widening of the TM6 linker to form a hydrophobic cavity that can interact with the diphenyl aromatic moieties of the two inhibitors of GABA transport, used in this study. Rather interestingly, the disordered region of TM6 reveals yet another modification to facilitate the formation of a new subsite to accommodate GAT1 inhibitors.

Figure 7. Altered subsite organization and substrate selection at the extracellular vestibule.

• A
  Altered subsite architecture is responsible for the altered inhibitor interactions within the GABA transporter.
• B
  Selective neurotransmitter entry in GABA and biogenic amine transporters is aided by EL4 residues.

The TM6 linker has displayed multiple attributes to regulate the volume and plasticity of the primary binding site in response to substrate and inhibitor interactions observed in earlier studies (Wang et al, 2015). In bacterial homologs of SLC6 members like LeuT and MhsT, the residues equivalent to Phe325 (Phe259^LeuT and Met236^MhsT) are observed to serve as volumetric sensors that adapt to diverse amino acids binding to the primary binding site (Singh et al, 2008; LeVine et al, 2019; Focht et al, 2021). In the current study, the F325L GAT1‐like substitution is in a subtly altered position that avoids clashes with the inhibitor aromatic groups that are ensconced within the TM6 linker in the newly identified subsite C′. This interaction was largely missed by most docking/modeling studies of GAT1 inhibitors with the diaryl moieties generally observed to be docked in the vestibule without completely entering the primary binding site (Skovstrup et al, 2010; Jurik et al, 2015).

Most antidepressants and psychostimulants that competitively inhibit biogenic amine uptake have interactions with subsite B wherein aromatic groups interact and enhance the affinity of inhibitors (Wang et al, 2015; Coleman et al, 2016; Pidathala et al, 2021). In the case of GAT1, the polar substitutions in the site create a H‐bond network around Thr425 (Thr400^GAT1) that disallows any inhibitor interactions in this pocket leading to a major reorganization of the orthosteric binding site.

Despite major alterations within subsites B and C, subsite A remains an important site for substrate recognition in all SLC6 neurotransmitter transporters. In biogenic amine transporters, aspartate residues in TM1b (Asp46^dDAT and Asp98^hSERT) are involved in recognition of primary and secondary amines in substrates and inhibitors, respectively. The residue is substituted to a glycine in GATs, GlyTs, and bacterial amino acid transporters (Fig EV1). The substitution allows the carboxylate group of the substrate to interact with the Na⁺ ion at site1 as observed in the NO711 and SKF89976a complexed structures, in this study.

Interestingly, both the inhibitors NO711 and SKF89976a that display high‐affinity, sub‐micromolar interactions with GAT1 interact with GAT2, GAT3, and BGT1 with significantly weakened affinities (Borden, 1996). The IC50 values for GAT2, GAT3, and BGT1 are 740, 350, and 3,570 μM, respectively, for NO711. SKF89976a displays IC50 values 550, 944, and 7,210 μM against GAT2, GAT3, and BGT1, respectively. Both the inhibitors interact at the primary binding site of GAT1 that has substantial discrepancy in the side chain environment in comparison with GAT2, GAT3, and BGT1 that are closer to each other with similar residue environment in the binding pocket in comparison with GAT1 (Appendix Fig S5). In the region of subsite A, the Tyr60^GAT1 (Phe43^dDAT) forms the base of the primary binding site. This residue is substituted to a glutamate in GAT2, GAT3, and BGT1 (Glu48^GAT2, Glu66^GAT3, and Glu52^BGT1) that substantially alters the charge of the binding pocket and local interactions of –OH group of Tyr60 that forms H‐bonds with carbonyl group of TM6 linker residue Gly297^GAT1 and the hydroxyl side chain of Ser302^GAT1. Substitution of Tyr60^GAT1 to glutamate in GAT1 and the alteration of the Glu48^GAT2 to a tyrosine in GAT2 led to highly compromised GABA uptake activity (Kanner, 2003; Schlessinger et al, 2012). The aromatic residues at this position, in GAT1 and other biogenic transporters, also form aromatic stacking interactions with the inhibitors that bind in the primary binding site. The substitution to a glutamate in GAT2 and GAT3 disrupts the aforementioned interactions and creates a local negative charge that may hamper carboxylate group containing inhibitors like tiagabine, NO711, and SKF89976a (Latka et al, 2020b). Further modifications include the presence of a glycine (Gly297^GAT1) substituted to alanine in the TM6a of GAT2 and GAT3 (Ala291^GAT2, Ala311^GAT3, and Ala296^BGT1) that constrict the space for nipecotic acid binding in comparison with GAT1. We also observe multiple substitutions of GAT1 residues in the binding pocket to cysteine that include Gly299^GAT1 (substituted to Cys293^GAT2, Cys313^GAT3, and Cys298^BGT1); Ser302^GAT1 (equivalent to Cys296^GAT2, Cys316^GAT3, and Cys301^BGT1); and Thr400^GAT1 (equivalent to Cys394^GAT2, Cys414^GAT3, and Cys399^BGT1). Although the models do not indicate the presence of a disulfide bond among the cysteine residues in the binding sites of GAT2, GAT3, and BGT1, future structural studies on these isoforms may reveal disulfide cross‐links within the binding pocket. These substitutions significantly alter the polar and hydrophobic interactions of the inhibitor binding site and disrupt the H‐bond networks formed by the residues in the direct vicinity of the inhibitor, likely leading to weakened interactions of GAT1‐specific inhibitors.

The complex of dDAT_GAT with SKF89976a provided a serendipitous observation of the inhibitor bound in the allosteric site in addition to the orthosteric site. Unlike the primary binding site in dDAT_GAT that carries 10 GAT1‐like substitutions, the extracellular vestibule was left largely unaltered and a single substitution was done in EL4 (E384S). Allosteric sites modulating neurotransmitter transport have gained prominence in the recent past with the identification of citalopram, vilazodone binding in hSERT at the allosteric site in the vestibule (Coleman et al, 2016; Plenge et al, 2021). Diverse interactions of noncompetitive inhibitors were also mapped in GlyTs with cmpd1 interacting with an inward‐open state and bioactive lipids proposed to interact with GlyT2 (Mostyn et al, 2019; Shahsavar et al, 2021). Although allosteric inhibitors of GABA uptake have also been characterized using mass spectroscopy‐based tools, their exact site of interaction is unknown due to the absence of a structure for any GAT isoform (Hauke et al, 2018). The SKF89976a bound in the allosteric site in the extracellular vestibule of dDAT_GAT can, therefore, aid in understanding allosteric inhibitors of GAT1. This is particularly important as SKF89976a was previously known to interact with GAT1 in both a competitive and noncompetitive manner and differentiate between GABA transport current and GABA evoked inward Na⁺ currents observed in GATs (Cammack & Schwartz, 1996; Eckstein‐Ludwig et al, 1999; Krause & Schwarz, 2005).

The interactions of the nipecotic acid moiety of SKF89976a, in the allosteric site with the E384S substitution, were investigated, and the results indicate that residues in EL4 region could be an additional site of control of neurotransmitter entry in dDAT but less so in GAT1. The negative charge of Glu384^dDAT in EL4 could also aid in attracting the positively charged monoamines to enter the vestibule. In DAT and SERT, the rate‐limiting step of substrate/inhibitor selectivity is observed to be the association rate (Hasenhuetl et al, 2015). EL4, due to its unique position to plug the extracellular vestibule can affect association rates and thereby substrate/inhibitor interactions within the transporter. Evidence of this is visible in EL4 of hSERT where substitutions (L406E) are observed to weaken substrate transport and allow competitive inhibitors to interact with increased potency through stabilizing the outward‐open state (Rannversson et al, 2015). However, our observations with equivalent EL4 substitutions in dDAT and GAT1 indicate that the two transporters behave rather distinctly in response to changes in EL4 with GAT1 showing a modest effect on substrate interactions and transport activity, whereas the dDAT displaying complete loss of dopamine uptake (Fig 7B). Insights into these distinct substrate translocation effects within the extracellular vestibule and the details of allosteric inhibitor interactions could be provided through future structural studies of GAT isoforms.

While an engineered construct of dDAT_GAT may not be a perfect substitute to study the diverse aspects of GAT1 structure and function, we could successfully translocate and mimic a GAT1 binding site into a catecholamine transporter, dDAT, to unravel some unknown facets of GAT pharmacology. The structures of GAT inhibitors complexed to dDAT_GAT can serve as templates for improved designs of allosteric and orthosteric modulators of GAT activity. This is particularly significant in the context of GABA transport inhibition being a major strategy for treatment of epileptic conditions like partial seizures.

Materials and Methods

Constructs used in the study

The Drosophila melanogaster dopamine transporter (dDAT) construct used for dDAT_GAT engineering (dDAT TS2) contains a deletion of 20 amino acids in the N terminus (Δ1–20), a deletion in the extracellular loop 2 (EL2) from 164 to 191 amino acids (Δ164–191) and also contains two thermostabilizing mutations, V74A and L415A. The dDAT_GAT construct used for crystallization studies consists of 11 substitutions that include F43Y, D46G, A117S, V120L, D121N, F325L, V327S, E384S, S422Q, G425T, and S426V. In addition to these 11 mutations, the construct has deletions Δ1–20 and Δ162–202 and thermostabilizing mutations V74A, V275A, V311A, L415A, and G538L. Residues 602–607 were replaced with a thrombin site (LVPRGS) toward the C terminus. All the dDAT mutants generated for dDAT_GAT engineering are mentioned in Appendix Table S1.

The dDAT_mfc construct contains five thermostabilizing mutations, V74A, V275A, V311A, L415A, and G538L and deletions Δ1–20 and Δ162–202. A thrombin site (LVPRGS) was inserted by replacing residues from 602 to 607.

The Rattus norvegicus GABA transporter 1 (GAT1 _WT ) gene used for the biochemical studies was subcloned into pEG‐BacMam vector between EcoR1 and Not1 restriction sites from pBluescript II vector.

Purification of dDAT_GAT

The dDAT_GAT construct was expressed as a C‐terminal GFP fusion protein in HEK293 GnTI⁻ cell lines using baculovirus‐mediated expression system (Goehring et al, 2014). Membranes were solubilized in 20 mM dodecyl β‐D‐maltoside (DDM; Anatrace), 2 mM cholesteryl hemisuccinate (CHS; Anatrace), 50 mM Tris‐Cl pH 8.0, 150 mM NaCl, and 0.1 mM PMSF by incubating for 2 h. The solubilized material was separated by ultracentrifugation at 100,000 × g for 90 min. The supernatant was affinity‐purified using Talon resin (pre‐equilibrated with 1× TBS) in 20 mM Tris‐Cl pH 8.0, 300 mM NaCl, 100 mM imidazole, 1% glycerol buffer containing 1 mM DDM, and 0.2 mM CHS. The affinity‐purified dDAT_GAT protein was treated with thrombin (Haematologic Technologies Inc) for 36 h at 4°C in order to cleave the GFP‐8x‐His tag at the C‐terminal. The cleaved protein was concentrated to 4 mg/ml and injected into a Superdex‐200 10/300 increase column (GE Life Sciences), which was pre‐equilibrated with size‐exclusion chromatography (SEC) buffer containing 20 mM Tris‐Cl pH 8.0, 300 mM NaCl, 5% Glycerol, 4 mM decyl‐β‐D‐maltoside (DM), 0.3 mM CHS, and 0.01% (w/v) 1‐pamitoyl‐2‐oleoyl‐sn‐glycero‐3‐phospho ethanolamine (POPE; Avanti Polar Lipids).

Expression and purification of Fab

Heterologous expression of the Fab 9D5 was carried out using baculovirus‐mediated insect cell expression system (Pidathala et al, 2021; Appendix Fig S2D). Fab heavy and light chain genes were synthesized (Genscript) and cloned into pFastBac Dual vector. The heavy chain is under the polyhedrin promoter, and the light chain is under the P10 promoter along with N‐terminal GP64 signal sequence on each chain. Sf9 cells in suspension media at a density of 2.5 × 10⁶ cells/ml were infected using high titer recombinant baculovirus, and the supernatant containing Fab was dialyzed against 25 mM Tris‐Cl (pH 8.0) and 50 mM NaCl. The Fab (9D5) was affinity‐purified from the dialysate using metal ion affinity chromatography in 25 mM Tris‐Cl (pH 8.0), 50 mM NaCl, and 250 mM imidazole. The affinity‐purified Fab was further purified using size‐exclusion chromatography by injecting into a superdex 75 10/300 GL column (GE Life Sciences) in 25 mM Tris‐Cl (pH 8.0) and 50 mM NaCl.

Crystallization and structure determination

Purified dDAT_GAT was incubated with varying concentrations of GAT1 inhibitors, NO711 and SKF89976a for 3 h at 4°C on prior to the addition of Fab 9D5 in a molar ratio of 1:1.2. The dDAT_GAT‐9D5 complex was incubated on ice for 30 min and concentrated to a final concentration of 3.0–3.5 mg/ml using a 100 kDa cutoff centrifugal concentrator (Amicon Ultra). The concentrated protein complex was ultra‐centrifuged at 100,000 × g for 60 min and subjected to crystallization using hanging drop vapor diffusion method at 4°C. The dDAT_GAT substrate‐free form and dDAT_GAT‐NO‐711‐bound crystals were obtained in 0.1 M MOPS pH 6.5–7.0 and 32–34% PEG 600, whereas dDAT_GAT‐SKF89976a‐bound crystals were obtained in 0.1 M Tris‐Bicine pH 8.0–8.2 and 32–34% PEG 600. Cocrystallization of inhibitors with dDAT_GAT was performed in the presence of 0.5 mM of both inhibitors. X‐ray diffraction data sets were collected at 24IDC at APS and XRD2, Elettra and crystals diffracted in a resolution range of 2.8–3.2 Å (Appendix Table S3). X‐ray data were processed using XDS, and scaling and merging were performed using AIMLESS in the CCP4 software suite (Kabsch, 2010; Evans & Murshudov, 2013). The dDAT_GAT structures were solved using molecular replacement by PHASER with the dDAT and 9D5 coordinates from PDB id 4XNU (McCoy et al, 2007). After fitting the ligands in the omit densities, multiple rounds of refinement were carried out using Phenix.refine and COOT in the PHENIX crystallographic software suite, respectively (Adams et al, 2010). Ligand fitting was done in COOT followed by real‐space refinement after fitting the inhibitor generated using SMILES. The real‐space correlation coefficient for the inhibitors in the dDAT_GAT structures is above 0.85.

GABA uptake assay

The uptake assays were performed using HEK 293 GnTI⁻ cells expressing respective constructs. Cells were resuspended in 25 mM HEPES‐Tris pH 7.1, 130 mM NaCl, 1 mM MgSO₄, 5 mM KCl, 1 mM CaCl₂, and 5 mM D‐Glucose containing uptake assay buffer. Control samples were pre‐incubated with 10 μM of tiagabine for 30 min before addition of radiolabel. One micrometre of GABA in a 1:200 molar ratio of [³H]‐GABA: [¹H]‐GABA (Perkin Elmer) was added and incubated for 30 min at room temperature. The reaction was arrested by the addition of ice‐cold assay buffer, and the cells were washed twice with the same buffer before solubilizing in 20 mM DDM. The solubilized sample was separated using centrifugation at 14,500 × g for 30 min, and the supernatant was added to 0.5 ml of scintillation fluid (Ultima Gold, Perkin Elmer). The radioactive counts were measured by liquid scintillation counting using MicroBeta scintillation counter (Perkin Elmer).

GABA uptake and inhibition assay in 96‐well plate format

For the determination of K _M , K _i , and V _max values, the GABA uptake assays were carried out in 96‐well plates (Aggarwal & Mortensen, 2017). HEK 293 GnTI⁻ cells were infected with the baculovirus of respective mutant, and cells were plated at a cell density of 50,000 cells per well 36‐h post‐infection. Uninfected cells were used as the control. The media were aspirated after 4 h, and the cells were washed with assay buffer containing 25 mM HEPES‐Tris pH 7.1, 130 mM NaCl, 1 mM MgSO₄, 5 mM KCl, 1 mM CaCl₂, and 5 mM D‐Glucose. The cells were incubated with varying concentrations of GABA (0.5, 1, 2, 4, 8, 10, 20, and 40 μM) in 1:500 molar ratios of [³H‐GABA] and [¹H GABA] and incubated for 15 min at room temperature. GABA uptake was arrested with 200 μl of ice‐cold assay buffer and washed twice with the same buffer before solubilization. Fifty microlitre of scintillation fluid was added to the solubilized material, and the radioactivity was estimated by scintillation counting using MicroBeta liquid scintillation counter. The background‐subtracted initial uptake rates were plotted against the above‐mentioned concentrations of GABA to determine the K _M and V _max values. The data were plotted and analyzed using GraphPad Prism v.5.0.1.

For the estimation of K _i value, we carried out GABA uptake assays in 96‐well plate format as described above. The cells were incubated in the uptake assay buffer containing varying concentrations of inhibitors for 30 min. This was followed by the addition of 8 μM GABA in 1:500 molar ratios of [³H‐GABA] and [¹H GABA] and incubated for 15 min at room temperature. The reaction was arrested with the ice‐cold assay buffer and washed with the same buffer before solubilization. Fifty microlitre of scintillation fluid was added to the solubilized material, and the radioactive counts were estimated on MicroBeta liquid scintillation counter. The background‐subtracted, dose–response plots were analyzed and plotted using GraphPad Prism v.5.0.1, and K _i values were determined from Cheng‐Prusoff's equation using the IC₅₀ values obtained from the experiments.

Dopamine uptake assay

The assay was carried out with the wild‐type and mutant dopamine transporter for the determination of K _M and V _max values using the uptake assay (96‐well plate format) described in the previous section. Dopamine uptake was initiated by the addition of varying concentrations of DA (0.5, 1, 2, 4, 8, 10, and 20 μM) in 1:200 molar ratios of [³H]‐DA: [¹H]‐DA at room temperature. The uptake was arrested after 5 min with the ice‐cold uptake buffer containing 200 μl of 25 mM HEPES‐Tris pH 7.1, 130 mM NaCl, 1 mM MgSO₄, 5 mM KCl, 1 mM CaCl₂, 5 mM D‐Glucose, 1 mM ascorbic acid, and 30 μM pargyline, and the cells were washed twice with the same buffer before solubilization. Fifty microlitre of scintillation fluid was added to the solubilized material, and the radioactivity was estimated on MicroBeta scintillation counter. The background‐subtracted initial uptake rates were plotted against the respective DA concentration, and values of K _M were determined.

Binding assay

Binding assay for the dDAT_GAT construct was carried out using microscale thermophoresis (MST; Nanotemper; Clémençon et al, 2018). 100 nM of SEC purified dDAT_GAT was labeled with 100 nM red Tris‐NTA dye. The labeled protein was incubated with 16 twofold serial dilution of the ligand keeping the protein concentration constant at 10 nM. Monolith™ NT.115 MST premium‐coated capillaries were used in all the experiments. The starting concentrations of ligands were 2 mM and dissolved in the same buffer as that of the protein (50 mM HEPES pH 8.0, 300 mM NaCl, 5% Glycerol, 4 mM DM, 0.3 mM CHS, and 0.01% (w/v) 1‐pamitoyl‐2‐oleoyl‐sn‐glycero‐3‐phospho ethanolamine (POPE)). The thermophoresis binding data were analyzed and plotted using Graphpad Prism. The dDAT_mfc construct was used as the control for MST binding assays.

Homology modeling

The homology models of GAT isoforms GAT2, GAT3, and BGT1 were built using the dDAT_GAT structures as templates in the swiss‐model online server (https://swissmodel.expasy.org/). The resulting models had very close overlap on the template structure with all greater than 98% of residues in the favored regions of the Ramachandran plot. The human GlyT1 outward‐open model used for structural comparison was obtained from the AlphaFold2 database (https://alphafold.ebi.ac.uk/; Varadi et al, 2022).

Data statistics

All the radiolabel biochemical assays were performed as two independent measurements each done in triplicate. All six measurements of individual data points were used to calculate the transport and inhibition parameters in the study using GraphPad Prism v.5.0.1. The error bars in all cases represent s.e.m, and unpaired two‐tailed t‐tests were performed to calculate significance values.

Author contributions

Deepthi Joseph: Conceptualization; Data curation; Formal analysis; Validation; Investigation; Visualization; Methodology; Writing—original draft; Writing—review and editing. Smruti Ranjan Nayak: Formal analysis; Visualization; Writing—review and editing. Aravind Penmatsa: Conceptualization; Resources; Data curation; Software; Formal analysis; Supervision; Funding acquisition; Validation. Investigation; Visualization; Methodology; Writing—original draft; Project administration; Writing—review and editing.

In addition to the CRediT author contributions listed above, the contributions in detail are:

DJ performed protein expression, purification, crystallization and biochemical measurements. DJ and SRN performed the heterologous expression of the antibody fragment. SRN aided in the preparation of figures. AP designed the study and performed crystallographic analyses. AP wrote the manuscript with inputs from all the authors.

Disclosure and competing interests statement

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Appendix

Source Data for Figure 1

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

The EMBO Journal (2022) 41: e110735.

Data availability

The coordinates for the structures have been deposited in the Protein Data Bank with the following accession codes 7WGD, 7WGT, and 7WLW (Appendix Table S3). The raw data and gel images for all biochemical experiments have been deposited alongside the manuscript as supplementary information and a source data file. Diffraction data would be made available upon request.