Transport mechanism of a glutamate transporter homologue Glt_Ph

Glutamate transporters, also known as excitatory amino acid transporters (EAATs), belong to the dicarboxylate/amino acid:cation (Na⁺ or H⁺) symporter (DAACS) family [1]. In the mammalian central nervous system, neuronal and glial EAATs transport glutamate, the main neurotransmitter, from the outside to the inside of the nerve cells, removing excessive excitotoxic glutamate, which may cause neurotoxicity [2,3]. Various human diseases, such as Alzheimer's disease, epilepsy and strokes, have been linked to dysfunction of EAATs [4,5].

In humans, there are five subtypes of glutamate transporters (EAAT1–5) [6]. The transport of glutamate is driven by energy derived from ion gradients, mostly Na⁺ [2,3,6]. In EAATs, three Na⁺ ions and one proton are co-transported with glutamate and the transport cycle is completed by the counter-transport of one K⁺ ion [7]. In addition to the ion-coupled transport, EAATs also display uncoupled chloride conductance [8–11] and have different preferences towards ions [10]. Therefore, glutamate transporters function both as secondary active transporters and anion-selective ion channels [8,10,12].

Despite the importance of glutamate transporters in mammalian systems, there are currently no crystal structures of a mammalian EAAT. One archaeal homologue of the glutamate transporter, Glt_Ph, isolated from Pyrococcus horikoshii glutamate transporter, has however been extensively studied over the past ten years. It shares 37% sequence identity with human EAAT2 [13,14] and many functionally important amino acid residues are highly conserved between Glt_Ph and its human homologues [13], making it an excellent model system for researchers to use.

Glt_Ph transports aspartate together with three Na⁺ ions into the cytoplasm [15], accompanied by a stoichiometrically uncoupled Cl⁻ conductance as well [16]. There are thus three major differences between it and the human EAATs: first that no proton is symported with aspartate [17], second that K⁺ ion counter-transport is not required to complete the transport cycle [17] and third, a strong preference for aspartate over glutamate [18]. In contrast, EAATs require one proton for co-transport [7], one K⁺ ion counter-transport to complete the transport cycle [7] and transport glutamate and aspartate with similar affinity [8,11,19]. In this review, we summarize the current state of structural studies, MD simulations and single-molecule FRET (smFRET) studies of Glt_Ph that have provided insights into its transport mechanism–and by extension, the mechanism of the EAATs as well.

The outward-facing state, captured in the first crystal structure of Glt_Ph [13], revealed a homotrimer (Figure 1a) with a bowl-shaped extracellular-facing basin whose surface is hydrophilic and as deep as half of the trimer's height. Each wedge-shaped protomer (Figure 1b) consists of two domains: a trimerization domain formed by four transmembrane (TM) helices (TM1, TM2, TM4 and TM5) providing interactions between subunits in the trimer; and a transport domain formed by four TM helices (TM3, TM6, TM7 and TM8) and two re-entrant loops [helical hairpin (HP) structures, HP1–2] [13,20].

Comparison of the aspartate-bound structure and the structure with the competitive inhibitor DL-threo-β-benzyloxyaspartate (TBOA)-bound shows that HP2 serves as the extracellular gate [18]. Glt_Ph can adopt an ‘open’ conformation (solved with TBOA bound), which allows substrate access from the outside to its binding site, at which point it switches to the ‘closed’ conformation (solve with aspartate bound). This role of HP2 has also been verified by MD studies [21,22]. HP1 was therefore proposed to function as the intracellular gate as its movement is involved in the dissociation and release into the cytoplasm of the substrate and ions [20]. However, this remains the subject of some controversy in recent MD studies, as will be discussed below (Transport Mechanism).

As the substrate-binding site in both the aspartate- and TBOA-bound structures is approximately 5 Å (1 Å=0.1 nm) beneath the extracellular surface, these two structures are called [20] the outward-facing closed (or occluded) state and outward-facing open state respectively. The inward-facing state is obtained by cross-linking of a double-cysteine mutant introduced into Glt_Ph [20] (Table 1). For example in the structure of Glt_Ph-K55C–A364C_Hg, aspartate is bound approximately 5 Å beneath the intracellular surface [20].

Biochemical, crystallographic and double electron–electron spin resonance [DEER (also called PELDOR)] spectroscopy data all demonstrate that the trimerization domain serves as a scaffold and stays in almost the same conformation during ligand binding and transport [20,23,24], whereas the transport domain, stabilized by the scaffold, undergoes large conformational changes involving a TM translation and rotation [20]. Various studies with different techniques performed on EAATs show that individual subunits in the homotrimer function independently [25–28]. Although there is no direct evidence about how the subunits in Glt_Ph function, it should be similar to the EEATs, given the high level of similarity between Glt_Ph and the EAATs.

Rigid body movement (called ‘elevator-like’ motions [29]) of the transport domain can be observed when comparing the structures of apo or holo outward-facing and inward-facing Glt_Ph respectively [20,30]. The elevator-like motions of the transport domain have also been observed in a smFRET study of the wild-type and a humanized mutant (R276S/ M395R) of Glt_Ph [31], suggesting that these motions mediate substrate uptake and are pivotal steps of the transport cycle [31,32].

Both DEER [24] and smFRET [32,33] studies on Glt_Ph show that the protomers in the trimer can sample different conformations randomly and independently, and individual transport domains alternate between periods of quiescence and periods of rapid transition. This is also captured in the Glt_Ph-V198C–A380C_Hg crystal structure, with one of the protomers in the intermediate outward-facing state and the other two in the inward-facing state [34].

The positions of two Na⁺ ions (Na1 and Na2) have been experimentally identified: there is no direct interaction between these two Na⁺ ions and the bound aspartate [18]. In the outward-facing holo crystal structure, Na1 is located below the aspartate, coordinated by the main chain carbonyls of Gly³⁰⁶ and Asn³¹⁰ (TM7), of Asn⁴⁰¹ (TM8) and the Asp⁴⁰⁵ side chain (TM8) (Figure 1c). Of these residues, Asp⁴⁰⁵ is the most important: it coordinates Na1 bidentately via the γ-carboxylate group, and analysis of data from the Glt_Ph–D405N crystals soaked in Tl⁺ solution (an Na⁺ mimic) found a strong peak only at the Na2, not the Na1, position and the mutant bound aspartate more weakly [18]. In the outward-facing holo crystal structure, Na2 is below the re-entrant helical HP2, coordinated by the carbonyl groups of Thr³⁰⁸ and Met³¹¹ (TM7) and of Ser³⁴⁹ and Thr³⁵² (HP2) [18] (Figure 1c).

In both the outward-facing and inward-facing holo crystal structures, the distance between the hydroxy group of Thr³⁰⁸ side chain and the backbone carbonyl of Pro³⁰⁴ is approximately 4.8 Å, which is too far to form a hydrogen bond. This allows Thr³⁰⁸ to coordinate the Na⁺ ion at Na2. However, the Pro³⁰⁴–Thr³⁰⁸ hydrogen bond exists in the outward-facing apo crystal structure [30] and the outward-facing crystal structure of Glt_Ph with TBOA bound [18]. In the outward-facing apo crystal structure, the HP2 loop is collapsed into the aspartate binding and Na2 sites as well [30]. In the structure of Glt_Ph in complex with TBOA, HP2 moves approximately 10 Å away from the position where it is in the outward-facing holo structure and therefore cannot coordinate an Na⁺ ion at Na2 [18]. Steered molecular dynamics (SMD) simulations suggested that the breaking of the hydrogen bond between Pro³⁰⁴ and Thr³⁰⁸ destabilizes the last turn of the TM7a helix and allows readjustment of the backbone carbonyl oxygen atoms, placing them in a favourable position to coordinate the second Na⁺ ion [35]: this role for Thr³⁰⁸ has been verified by experimentally measuring the involvement of three Thr³⁰⁸ mutants (T308W/T308A/T308V) in binding and transport [35]. However, superposition of the outward-facing apo Glt_Ph structure and the outward-facing Glt_Ph structure with Na⁺ bound at Na1 shows that ion binding to Na1 releases HP2 to free the aspartate binding and Na2 sites into a conformation similar to that in the outward-facing holo structure, breaking the hydrogen bond between Thr³⁰⁸ and Pro³⁰⁴ as well [30].

The third Na⁺-binding site (Na3) is difficult to observe structurally, because binding at the third site would lead to conformational change and transport. Consequently, opinions vary regarding its position [36–38]. An MD simulation [38] based on the aspartate-bound and TBOA-bound structures [18] predicted a new binding site for Na3, which differs from previous MD simulation results [36,37]. Bastug et al. [38] predicted that the third Na⁺ ion is coordinated by the side chains of Thr⁹², Ser⁹³, Asn³¹⁰, Asp³¹² and the backbone of Tyr⁸⁹. They were able to verify this experimentally: the T92A and S93A variants showed different changes in aspartate affinity but both exhibited a reduction in Na⁺ affinity compared with wild-type Glt_Ph. In addition, Asn³¹⁰ and Asp³¹² are both part of the highly conserved NMDGT motif [18]; Thr³¹⁴ in the motif is involved in aspartate binding [18] and mutations of the equivalent residue (Thr⁴⁰⁰) in EAAT2 abolish its function [39].

A stoichiometrically uncoupled Cl⁻ conductance is observed along with aspartate transport in Glt_Ph [16]. This Cl⁻ conductance can partially neutralize the membrane potential caused by the electrogenic substrate transport. The anion selectivity of Glt_Ph is almost the same as that of EAATs. Mutation of a conserved amino acid (S65V in Glt_Ph, located in TM2) strongly affects the chloride conductance with almost no effect on the Na⁺: aspartate symporter [16], similar to results observed in EAAT1 (S103V) [40]. Clearly, Cl⁻ permeates through a specific pathway [16] and Ser⁶⁵ is somehow involved in the process. In a recent MD simulation [41], however, researchers were unable to find any evidence showing that Ser⁶⁵ interacts directly with Cl⁻. Combined with experimental evidence obtained from both Glt_Ph and EAAT4, they proposed that Ser⁶⁵ exerts its effect on anion permeation by altering the rates of conformational changes leading to the open anion channel.

A recent study combined MD simulations with fluorescence spectroscopy of Glt_Ph and patch-clamp recordings of mammalian EAATs [41]. The authors suggested that lateral movement of the transport domain triggers formation of the anion-selective permeation pathway only if the domain sampled intermediate transporter conformations, rather than outward- or inward-facing states. They predicted residues that line the ion permeation pathway by simulation and verified these predictions through fluorescence spectroscopy and functional studies on mutant transporters. Of the residues lining the pathway, the side chain of Arg²⁷⁶ protrudes from the tip of HP1 into the Cl⁻ permeation pathway and this resulting positive charge contributes to the anion selectivity for both Glt_Ph and the EAATs [41]. This residue is also involved in the binding of substrates [18,30]. Interaction with the substrate does not compromise its role in anion permeation and selectivity [41].

Although Glt_Ph is a glutamate transporter homologue, it exhibits a strong preference for aspartate as a substrate in the presence of an Na⁺ gradient. It shows 60000-fold higher affinity for aspartate (with K_d values for aspartate and glutamate of approximately 2 nM and 122 μM respectively) [18]. The aspartate-binding site consists of the tips of HP1 and HP2, the conserved NMDGT motif of TM7 (see above) and hydrophilic residues on TM8 [18] (Figure 1d). The α-carboxyl group of the substrate interacts with the side chain of Asn⁴⁰¹ (TM8) and the main chain amide nitrogen of Ser²⁷⁸ (HP1), whereas the γ-carboxyl group interacts with the side chains of Thr³¹⁴ (TM7) and Arg³⁹⁷ (TM8). The substrate amino group interacts with the side chain of Asp³⁹⁴ (TM8) and the backbone carbonyl groups of Arg²⁷⁶ (HP1) and Val³⁵⁵ (HP2) (Figure 1d).

Binding thermodynamics studies show that aspartate binding and release, rather than TM movements of the transport domains, is coupled to the chemical potential of sodium ions in solution [42].

Structural comparison of outward-facing apo and holo-Glt_Ph shows that in the apo structure, there is joint movement of HP2 and TM8a and also reorganization of ligand-binding sites including HP2, the NMDGT motif and TM3. The HP2 loop region collapses into the substrate- and Na2-binding sites. The movements of side chains in the NMDGT motif (Asn³¹⁰ and Met³¹¹) and the bending away of TM3 from the motif deform the Na1 site [30]. These distortions mean than Na⁺ can no longer bind. (Similar distortion of ligand-binding sites also has been observed in the outward-facing apo structure of Glt_Tk [43], which has 77% sequence identity with Glt_Ph).

Binding of Na⁺ and aspartate trigger different movements of HP2, with the binding of the former causing HP2 to open and allow binding of Na2, whereas the binding of the latter causes HP2 to close [44]. Binding of aspartate and the Na⁺ at the Na2 site is coupled as both sites are partly formed by the tip of HP2 [30] (Figure 2). A binding thermodynamics study of Glt_Ph also suggests that binding of the first two Na⁺ is involved in the modification of the substrate-binding site, whereas the binding of the third Na⁺ is coupled to the substrate occlusion from outside solvent [42]. During the ligand binding process, with the exception of extracellular gate HP2 closure, other unknown conformational changes dominate the process and remain to be elucidated by further research [42]. After the ligands are fully bound to the transport domain and occluded from the solvent by the closure of both HP1 and HP2, the transport domain moves across the membrane as a rigid body [20] (Figure 2).

Simulations based on the inward-facing crystal structure of Glt_Ph have provided preliminary insights into the process of substrate release into the cytoplasm. DeChancie et al. [45] suggested that release is initiated by dissociation of Na⁺ from the Na2 site and, almost simultaneously, opening of the HP2 loop exposes the substrate and other polar and charged groups. This attracts water molecules to the substrate-binding site, which further destabilizes interactions between substrate and protein residues on HP2 and TM8. The HP1 loop then opens, disrupting the strong hydrogen bonds between the SSS motif (Ser²⁷⁷–Ser²⁷⁹) on the HP1 loop and the substrate, allowing the aspartate to dissociate. In this model, HP2 serves as an activator of the intracellular HP1 gate [45]. However, a previous simulation suggested that HP2 is in fact the intracellular gate in the inward-facing state [46]. In this model, HP2 opening is a prerequisite for substrate release into the cytoplasm. Understanding the mechanism of substrate release requires further research.

Following substrate release, the transport domain undergoes a series of conformational changes to prepare itself for the TM movement. The conformational changes in the inward-facing apo structure are that though all of the ligand-binding sites are distorted, the apo transport domain is as closed and compact as in the fully bound structure [30] (Figure 2). This may be critical for the transport domain to transit to the outward-facing state.

Although crystallographic, MD simulations and smFRET studies have greatly increased our understanding of the Glt_Ph transport mechanism, there are still many questions yet to be answered, including a definitive answer to the position of the third Na⁺ ion, the mechanism of substrate binding and release, and how the transport cycle is completed. Single-molecule and structural studies backed up by computational studies should yield definitive insights into the mechanism of substrate release and the transition to the outward-facing state in Glt_Ph. However, to understand the differences between it and the EEATs, for instance the differing substrate and ion transport specificity, will require high-resolution structures of the EEATs, either by X-ray crystallography or–possibly–by EM using the new generation of microscopes.

Adrian Goldman thanks Royal Society and Wolfson Foundation for Royal Society Wolfson Research Merit Award.

This work was supported by the China Scholarship Council (to Y.J.); the Wellcome Trust [grant number 019322/7/10/Z]; the Erkko Foundation [grant number 4704339] and the Biotechnology and Biological Sciences Research Council (BBSRC) [grant number BB/M021610/1] (all to A.G.) the Ministry of Science & Technology 973 Project [grant number 2014CB560709 (to M.B.)]; and the National Science Foundation of China [grant number 31322012 (to Y.W.)].

DEER

double electron–electron spin resonance

EAAT

excitatory amino acid transporter

Glt_Ph

Pyrococcus horikoshii glutamate transporter

HP

hairpin

smFRET

single-molecule FRET

PDB

protein data bank

TBOA

DL-threo-β-benzyloxyaspartate

TM

transmembrane