Glutamate transporters: a broad review of the most recent archaeal and human structures

Glutamate transporters play important roles in bacteria, archaea and eukaryotes. Their function in the mammalian central nervous system is essential for preventing excitotoxi-city, and their dysregulation is implicated in many diseases, such as epilepsy and Alzheimer ’ s. Elucidating their transport mechanism would further the understanding of these transporters and promote drug design as they provide compelling targets for understanding the pathophysiology of diseases and may have a direct role in the treatment of conditions involving glutamate excitotoxicity. This review outlines the insights into the transport cycle, uncoupled chloride conductance and modulation, as well as identifying areas that require further investigation.


Introduction
Glutamate is the most important excitatory transmitter in the brain. It carries messages between cells that are essential for normal development and function related to learning and memory. Additionally, glutamate molecules can promote connections among brain cells and remove unnecessary ones [1].
Glutamate transporters, also called Excitatory Amino Acid Transporters (EAATs), are members of the solute carrier 1 (SLC1) family of the secondary active transporters' important targets for drug development. SLC1 includes a range of human and prokaryotic glutamate and aspartate transporters, and two neutral amino acid transporters Alanine Serine Cysteine Transporters 1 and 2 (ASCT1 and 2) [1]. The five human glutamate transporters (EAAT1-5) are expressed abundantly throughout the central nervous system [2]. Glutamate transporters are responsible for ensuring that glutamate is constantly pumped into the cytoplasm, thus preventing cytotoxicity and neural damage. Eukaryotic glutamate transporters are the secondary active transporters as they couple the influx of one glutamate, three sodium ions and one proton with the efflux of one potassium ion [3] ( Figure 1A). In addition, eukaryotic glutamate transporters exhibit thermodynamically uncoupled chloride conductance resulting in an overall uncoupled glutamate-gated anion channel activity [4][5][6]. Archaeal transporter homologues are highly selective for aspartate over glutamate and, unlike EAATs, transport does not depend on proton or potassium gradients [7] ( Figure 1B). ASCTs, a structurally related family, mediate Na + -dependent exchange of small neutral amino acids at the plasma membrane [3]. Like the EAATs, archaeal homologues and ASCTs exhibit uncoupled anion conductance [8,9].
As the major excitatory transmitter, glutamate has a key role in the pathogenesis of different neurological disorders [1]: excess extracellular glutamate is toxic to the surrounding neurons via overstimulation that can result in cell damage and/or death. This has been linked to epilepsy, episodic ataxia, schizophrenia and Alzheimer's disease [10], and so EAATs have attracted much attention as they represent targets for understanding and treating these disorders [11].
The best structurally characterized members of SLC1 transporters are archaeal homologues Glt Ph and Glt Tk [12][13][14][15][16][17][18][19][20][21] (Table 1). They share ∼36% sequence identity with eukaryotic EAATs [22] with many conserved residues involved in substrate and ion binding and transport [13]. The progress in crystallographic studies of the glutamate transporter family, including the structure of human EAAT1 [23], has revealed much about the mechanism of glutamate transporters. The recent structure of glutamine transporter ASCT2 solved by cryo-EM [24] (Table 1), resembles of Glt Ph and Glt Tk and EAAT1 structures and reveals differences in substrate specificity as well as transport modes between ASCT2 and EAAT1.
This review summarizes the current understanding of the structure, transport and modulation mechanisms of glutamate transporters, highlighting what remains to be discovered and how that might be achieved.

Structure of glutamate transporters
Glutamate transporters are typically homotrimeric [13,25,26] (Figure 2), although EAAT3 and EAAT4 may also co-assemble as a heterotrimer [27]. The existence of a trimeric structure was first hypothesized on the basis of the archaeal glutamate transporter homologue Glt Ph [7,13], and the first crystal structure of the human EAAT1 [23] shown of a symmetric homotrimer in an outward-facing conformation (OFC) ( Table 1).
Each subunit ( protomer) of the trimer contains eight α-helical transmembrane segments (TM1-8) and two re-entrant loops, known as helical hairpins (HP1 and HP2). They are arranged into a peripheral 'transport domain' and a central trimerization 'scaffold domain' ( Figure 2C). The latter is formed by TM1, TM2, TM4 and TM5 [12,17]. Three scaffold domains form a central structure of the trimer and embed it in the membrane.
TM3, TM6, TM7, TM8 and re-entrant helical hairpins HP1 and HP2 form the peripheral transport domain ( Figure 2C). It contains motifs involved in substrate and cation binding. HP1 reaches up to the bottom of the extracellular basin, whereas HP2 is situated almost parallel to the membrane plane, with a large fraction exposed to solvent facing the extracellular basin in the OFC [13]. HP2 controls substrate and ion access to their binding sites [23].
The recently solved cryo-EM structure of ASCT2 [24] is highly similar to structures of other SLC1 members, although EAAT1 and ASCT2 scaffold domain have different orientations relative to the scaffold domain [24].

Sodium-and aspartate-binding sites
The structures of holo and apo Glt Tk allowed the identification of the sodium-and substrate-binding sites (Table 1) [20]. Na1, on the cytoplasmic side below the transported aspartate, is stabilized by the conserved Asp409 side chain on TM8 and main-chain carbonyl groups from the central region of TM7 (Gly309, Asn313) and TM8 (Asn405) ( Figure 3C). Na2 is just under the re-entrant hairpin motif HP2 (above the transported aspartate, nearer the extracellular side) and is coordinated by the main chain carbonyls of TM7 (Met314) and of HP2 (Ser352, Ile353, Thr355) ( Figure 3D) connecting TM7 and TM8. Na3 ( Figure 3E) is between the (A) The transport of one glutamate (red) is coupled to the co-transport of three Na + ( purple) and one proton (blue), and the counter-transport of one K + in the case of the human transport cycle. (B) The archaeal transport cycle is similar but involves the coupled transport of one aspartate (red) and 3 Na + ( purple) with no counter-transport. One of the intermediate transport   [20] and is coordinated by hydroxyl groups of Thr94 and Ser95 from TM3, the carboxamide group of conserved Asn313 and the sidechain carboxyl of Asp315 and the main chain carbonyl of Tyr91. Despite the names, the order of affinities is Na3 > Na1 > Na2, based on structural and thermodynamic studies [28][29][30]. For instance, the interaction distances appear to be the shortest for Na3 (∼2.2-2.4 Å, in comparison with ∼2.5-2.7 Å for Na1 and Na2), indicating that it binds with the highest affinity. Free energy calculations performed in the OFC for the Glt Ph support the assignment of Na3 as the tightest-binding cation [30]. Simulations also suggest that Na2 binds last, after substrate, and has the weakest affinity [29].
The residues in the substrate-binding site of eukaryotic glutamate transporters and archaeal transporter homologues are mostly conserved [12,15,16,23] (Figure 3A,B). The aspartate, situated in the centre of the membrane plane in the OFC, is coordinated by the tips of HP1 and HP2, and side chains on TM7 and TM8 [15]. Thr317 (TM7) interacts with the γ-carboxyl group of the aspartate, the side chain of Asn405 (TM8) and the main chain of Ser280 (HP1) interact with the α-carboxyl group, and the amino group of the aspartate forms contact with the side chain of Asp398 (TM8) and the main chain carbonyl groups of Val358 (HP2) and Arg278 (HP1) ( Figure 3A).  Crystal structures of human members of the glutamate transporter family [23,24] showed high similarities among homologues. A crystal structure of an aspartate-bound thermally stabilized mutant of EAAT1 revealed a substrate-binding site nearly identical with that of Glt Ph [23]. The conserved residues on TM8 (Asp456, Arg459 and Asn463) and TM7 (Thr382) coordinate the aspartate identically to the corresponding archaeal amino acids. Despite some differences in amino acids identity on HP1 (Ser343 and Ser345 (human) vs Arg278 and Ser280 (archaea) and HP2 (Ile423 (human) vs Val358 (archaea)), coordination of aspartate via the backbone atoms of these residues is maintained ( Figure 3A,B).
Comparison of the glutamine-binding site in ASCT2 and the glutamate-binding site in EAAT1 revealed the structural basis for the differences in the selectivity and binding mechanism of these highly related transporters. The majority of the residues in the binding sites are conserved (Asp464 (456), Asp460 (452), Thr468 (460), Ser353(345), Asn471(463) (EAAT1 numbering in brackets)), but the location of Cys467 in ASCT2 is occupied by Arg459 in EAAT1 and Ala390 is replaced by Thr382 [24]. Site-directed mutagenesis has shown that Cys to Arg substitution alters the selectivity of these transporters from neutral to charged amino acid substrates [31]. Thr382 is essential for the cooperativity of Na + and glutamate binding in the concentrative transporter EAAT1, but it has been suggested that this functionality may not be needed in the equilibrative ASCT2 [24].

Cooperative binding mechanism between sodium ions and aspartate
Glutamate transporters have a coupled binding mechanism: the chemical potential of sodium ions is coupled to the binding and release of the substrate. A recent kinetic study described cooperative binding as a key mechanism contributing to coupled transport by Glt Ph [32]. Consistent with a cooperative binding, substrate binding is fast and of high affinity [33] (K d of 2 nM) in the OFC when the extracellular concentration of Na + is high (200 mM) [12]. In the absence of substrate, HP2 adopts an open conformation, making the substrate-binding site accessible to the extracellular solution [12]. The tip of HP2 is open in the crystal structure of sodium-bound OFC Glt Ph [19], thus preventing uptake of uncoupled sodium ions [12].
Local rearrangements explain the coupling mechanism as shown by binding studies [18,34] and molecular dynamics simulations [29]. The slow and weak binding of two sodium ions [33,34] is associated with transporter core remodelling into a state that enables it to bind substrate and the remaining sodium ion, as shown by a cross-linking approach that trapped the transporter in key conformational states [18]. Aspartate does not bind with measurable affinity in the absence of Na + [18]. Kinetic studies confirmed that sodium ion binding occurs before substrate binding, but both one-sodium and two-sodium mechanisms appeared possible [32].
The most striking difference between the apo (ligand-free) and holo (ligand-bound) OFC structures is the conformation of the conserved NMDGT motif in TM7 [20]. Sodium binding to Asn313 in the Na3 site changes the geometry of the Na1 site to allow another sodium ion to bind, stabilizing the region. Gly316 and Thr317 in the NMDGT motif move closer to TM8 [20]. This, and the Na3 binding-mediated movement of TM3 towards TM7, places Thr317 in the optimal position to bind aspartate and reorients the side chain of Arg401, creating space for the aspartate to bind and allowing the guanidium group to interact with the β-carboxylate of aspartate. The interlinked Na3 and Na1 binding thus increases the affinity for aspartate. Na2 is likely the last site to fill [12], and its binding is coupled to the positioning of Met314 at the centre of the triangle formed by the bound substrate and two sodium ions. These structural changes provide compensatory interactions, allowing the HP2 tip to close [20].
There is no EAAT structure with glutamate bound. Kinetic studies on eukaryotic glutamate transporters show weak binding of 1 or 2 Na + ions prior to the binding of the substrate which is followed by the binding of the last Na + [35]. Although glutamate and aspartate binding should be similar, molecular dynamics (MD) simulations of homology-modelled EAAT3 suggest that glutamate is not stable in the amino-acid-binding site unless Glu374 is protonated [36]. The co-transported proton appears to be required for glutamate binding and transport. This needs further experimental investigation.

Transport motions
The transport cycle, driven by ionic gradients, involves alternate binding and release of the substrate on the extra-and intracellular sides of the plasma membrane due to the conformational transitions between OFC and IFC [18]. The alternate opening of the extracellular and intracellular gates is strictly maintained: in the OFC, when HP2 opens to expose the substrate and ion-binding sites to the extracellular solution, HP1 is secured in the closed state, packed against TM2 and TM5. Conversely, in the IFC, HP2, the extracellular gate, is locked closed upon displacing HP1 [17] (Figure 1).
Large rigid-body transport domain movements, relative to the static scaffold domain have been observed in the archaeal homologue of glutamate transporters for the first time [21]. The crystal structures of Glt Ph in OFC and IFC [12,13,17,19] showed that, during the transport cycle, the transport domains' elevator-like movements are supported by relatively rigid scaffold domains [17,21]. In the transition between OFC and IFC, the transport domain undergoes large transmembrane movements [37], involving a rotation of 37°and a translation of 15-18 Å that results in the substrate and ion-binding sites facing the cytoplasm [17], with rotation preceding translation. A complex motion between TM2 and TM4 as well as the conformational shift of TM4 may occur [38]. The scaffold domain, however, does not undergo conformational changes: its role is to mediate the intersubunit contacts in the transporter [18].
A single-molecule fluorescence resonance energy transfer study reported that the three Glt Ph subunits undergo independent conformational changes in an unsynchronized manner during aspartate transport [39] in accordance with the biochemical experiments conducted on eukaryotic transporters [40,41]. Furthermore, highspeed atomic force microscopy demonstrated that the protomers in the trimer act independently of each other: there was no correlation between the motions of individual transport domains [42]. Even though there were frequent transitions between OFC and IFC in the apo transporter, only one domain was in the IFC at one time [42]. When Na + was bound, there was a significant decrease in the fraction of moving protomers: protomers that bound only Na + were mostly unable to translocate. On the other hand, Glt Ph translocation dynamics increased when both Na + and aspartate were bound [42].
Molecular-level understanding requires comparison of the apo and holo OFC and IFC structures [21]. In the apo form, the transport domain has the same structure in both OFC and IFC [19]. However, the quality of the IFC structures is too poor for steered MD simulations and higher-resolution crystal structures of the intermediate forms are unlikely, due to the high heterogeneity of conformations and their short life-times. Recent developments in time-resolved electron microscopy [43] may help in this regard.

Substrate release
Once three Na + ions and aspartate are bound to the transporter, Na2 is in rapid equilibrium with a solution and its release is required for the disassembly of the complex [32]. Although the mechanism of substrate release is still somewhat unclear, Dechancie and colleagues [44] considered HP1 as the possible IFC-gate controlling element. They suggested that, following transition to the IFC, dissociation of Na2 triggers HP2 loop opening, therefore exposing and hydrating the aspartate and the other sodium ions and destabilizing the interactions between the substrate and the residues on HP2 and TM8. This causes opening of the HP1 loop, disrupting the hydrogen bonds between the Ser277and Ser279 (SSS) motif on HP1 and the substrate and allowing substrate release. As HP2 moves at least 4 Å away from the core, it may serve as the intracellular gate in the IFC [45]. Increase in the distance between TM8 and HP2 enables the substrate rearrangement across the space between HP1, HP2 and TM8. Furthermore, the interaction between the guanidium group of Arg276 at the HP1 and the β-carboxylate of the aspartate promotes the substrate translocation toward the cell interior. More recent Glt Tk structures with substrate and sodium ions bound showed that the loop between TM3 and TM4 ( Figure 2C), important in the transport, appeared to restrict HP2 movement in the OFC [20]. This may also occur in the IFC.
In ASCT2, the tip of HP2 is located on the cytoplasmic side of the membrane and, as it is not packed against the scaffold domain, HP2 in the IFC might possibly open in a similar manner as has been shown for Glt Ph and EAAT1 in the OFC [24]. Therefore, a single gate might alternately allow access from both sides of the membrane, without the need for HP1 opening [24].
Molecular dynamics and free energy calculations on the inward-facing conformation (IFC) of Glt Ph predicted that unbinding occurs as follows: Na2 → gate opening → Asp → Na1 → Na3 [28], with Na3 release being rate limiting. After ligand release, the apo transport domain adopts a compact, occluded conformation that can traverse the membrane, completing the transport cycle. Interactions between HP1, HP2, TM7 and TM8 stabilize the apo conformation [19]. Higher-resolution structures of the IFC will be required to determine the identity of the intracellular gate and mechanism of substrate release in more detail.

Potassium counter-transport
Human EAATs must counter-transport a K + to the extracellular medium for the reorientation from the IFC to the OFC [3]. Potassium binding to the IFC has been proposed to enable translocation of the substrate-free transport domain to the OFC [19]. Verdon and colleagues claimed to identify a potassium-binding site in Glt Ph [19], but subsequent structure analysis led to this being disregarded [22].
MD simulations on homology-modelled EAAT3 [36] suggested that a site close to the Na1 and Na3-binding sites (Figure 3) was the most likely as it had the highest affinity of the three considered. However, the K + ion was placed directly in the putative site, so the accessibility of binding sites and potential conformational changes between apo and holo forms was ignored. This is relevant because one of the sites was near several positively charged residues, which could prevent K + entering and binding. The simulations were also just 50 ns, so the ion position was not stable.
Confirmation of any of these binding sites requires a structure of the holo complex, which is not yet available [23]. Moreover, the domain motions following K + binding leading to the relocation to the OFC need to be defined.

Uncoupled chloride conductance
The chloride conduction pathway and the mechanism of anion permeation are still not completely understood. The anion selectivity of Glt Ph , the main subject of anion conductance studies, is almost identical with that of EAATs [46]. Chloride permeation occurs at the interface between the scaffold and transport domains, but it is not clear what conformational state mediates anion permeation nor exactly which residues are involved.
Initially, chloride permeation was proposed to occur along TM2 [47]. The Ser to Val mutation on TM2 (residue 65 in Glt Ph and 103 in EAAT1) strongly affected chloride permeation [8,47]. The Glt Ph structure in an intermediate OFC suggested that there was an aqueous cavity that might serve as a permeation pathway for anions [14]. Subsequent research has suggested that substrate transport and anion permeation proceed through two mutually exclusive pathways separated by a flexible wall domain and facilitated by the conformational changes [48,49]. The moving flexible wall is composed of several residues in HP1, HP2 hairpins and TM8 (Figure 4), the movement of which dictate the pathway [49]. Arg276 in HP1 and Met395, predicted to line the anion pore [49], were shown to exhibit a significant Cl − selectivity over Na + in both simulations and experiments: anion selectivity is impaired by the insertion of negatively charged side chains at specific positions. In contrast with previous research [48], mutations of Ser65 did not affect anion permeation. Presumably, therefore, Ser65 (Ser103) is part of the channel-opening mechanism but not of the permeation pathway. As the permeation pathway accounted for all known functional properties of EAAT/Glt Ph anion channels during simulations, it is the most likely mechanism of uncoupled chloride conductance.
Overall, it appears that both archaeal and mammalian transporters generate an open channel and permeation pathway through the cooperative dislocation of a flexible wall during the lateral movement of the transport core. The structural data obtained for anion conductance may be important for designing glutamate transporter anion channel modulators.

Modulation
Modulators are valuable tools for the study of the mechanism of glutamate transport [23]. Initially, the only known modulators were competitive substrate analogues like DL-threo-β-(benzyloxy)aspartic acid (TBOA) [12,23]. The extracellular gate was identified by a crystal structure of Glt Ph in complex with TBOA [12].
The structures of Glt Ph with aspartate and with TBOA are similar (Table 1), as is the conformation of the bound ligands ( Figure 5A) [12]. In the TBOA:Glt Ph crystal structure, HP2 is an OFC open conformation, moving from its position in the aspartate-bound complex towards the loop between TM3 and TM4 ( Figure 2C). This HP2 movement in the TBOA-bound complex exposes the substrate-binding site to the extracellular solution, suggesting that, in the apo state, HP2 adopts an open conformation before aspartate binding [22]. The structure of substrate-free transporter for the Glt Tk homologue revealed an OFC with occluded binding site and closed HP2 [15]. TBOA blocks HP2 in an open conformation, preventing sodium ion binding and subsequent conformational changes and thus inhibiting transport [12].
Selective, allosteric modulators of glutamate transporters are needed: such modulators have great potential to offer insights into the transport mechanism. The best-studied modulator is the non-competitive EAAT1-selective inhibitor UCPH 101 . The structure of thermostabilized EAAT1 in complex with L-aspartate and UCPH 101 showed that it bound at the scaffold domain-transport domain interface ( Figure 5B) [23], over 15 Å from the substrate and sodium-binding sites. UCPH 101 , therefore, does not preclude the movements of HP2 involved in binding of substrate and sodium ions from the extracellular solution. It does, however, induce local and global conformational changes to optimize its coordination in an OFC, thereby preventing transport [23]. A Glt Ph monomer viewed from the trimerization interface. The residues R276 and M395 responsible for the anion selectivity are indicated. The orange circles represent 'snapshots' of the path of a chloride ion through the channel between the two domains.
Part of the scaffold domain has been removed for clarity. Figure adapted from [49].
The structure of UCPH 101 has provided important insights into the conformational changes in EAAT1 that occur upon binding of substrate and sodium ions. Furthermore, it highlights the importance of allosteric cavities that can facilitate the design of novel, selective compoundsa feature common in membrane proteins [50] (Tate, pers. comm.).

Perspectives and future directions
• Mutagenesis, crystallographic and molecular dynamics studies have led to considerable progress in understanding glutamate transporters and their transport dynamics. Recent highresolution crystal structures of Glt Tk and EAAT1 and cryo-EM structure of ASCT2 have provided new insights into the transport mechanism.
• There are many areas that still require further investigation: glutamate, proton and K + binding are still not fully understood and the domain motions triggered by their binding need to be better defined. Our understanding of substrate release, and of the mechanism of potassium counter-transport, remains incomplete. The permeation pathway for chloride ions as well as its relation to substrate transport requires further validation.
• Improved IFC structures, either by X-ray crystallography or by (time-resolved) cryo-electron microscopy should lead to a better understanding. New modulators will further our knowledge of these transporters. We believe that future work will lead to molecules that allow the manipulation of transporter functions and, eventually, the treatment of pathological conditions.