Binding and transport of D-aspartate by the glutamate transporter homolog GltTk

Mammalian glutamate transporters are crucial players in neuronal communication as they perform neurotransmitter reuptake from the synaptic cleft. Besides L-glutamate and L-aspartate, they also recognize D-aspartate, which might participate in mammalian neurotransmission and/or neuromodulation. Much of the mechanistic insight in glutamate transport comes from studies of the archeal homologs GltPh from Pyrococcus horikoshii and GltTk from Thermococcus kodakarensis. Here, we show that GltTk transports D-aspartate with identical Na+: substrate coupling stoichiometry as L-aspartate, and that the affinities (Kd and Km) for the two substrates are similar. We determined a crystal structure of GltTk with bound D-aspartate at 2.8 Å resolution. Comparison of the L- and D-aspartate bound GltTk structures revealed that D-aspartate is accommodated with only minor rearrangements in the structure of the binding site. The structure explains how the geometrically different molecules L- and D-aspartate are recognized and transported by the protein in the same way.


Introduction
Mammalian excitatory amino acid transporters (EAATs) are responsible for clearing the neurotransmitter glutamate from the synaptic cleft (for review see Grewer et al., 2014;Takahashi et al., 2015;Vandenberg and Ryan, 2013). EAATs are secondary transporters that couple glutamate uptake to co-transport of three sodium ions and one proton and counter-transport of one potassium ion (Levy et al., 1998;Owe et al., 2006;Zerangue and Kavanaugh, 1996). EAATs transport L-glutamate, L-and D-aspartate with similar affinity (Arriza et al., 1994).
Here, we present the structure of Glt Tk with the enantiomeric substrate D-aspartate. The crystal structure was obtained in the outward-facing state with the substrate oriented in a very similar mode as L-aspartate, showing that the two enantiomers bind almost identically regardless of the mirrored spatial arrangement of functional groups around the chiral Ca atom.

Results
Affinity of D-aspartate and stoichiometry of sodium binding to Glt Tk Using Isothermal Titration Calorimetry (ITC), we determined the binding affinities of D-aspartate to Glt Tk in the presence of varying concentrations of sodium ions ( Figure 1A, Table 1). The affinity of the transporter for D-aspartate was strongly dependent on the concentration of sodium, similar to what has been reported for L-aspartate binding to Glt Ph and Glt Tk (Boudker et al., 2007;Hänelt et al., 2015;Jensen et al., 2013;Reyes et al., 2013). At high sodium concentration (500 mM), the K d values of Glt Tk for D-and L-aspartate binding level off to 374 ± 30 nM and 62 ± 3 nM, respectively. The DH values for binding of both substrates were favorable, with a more negative value of~1 kcal mol À1 for L-aspartate, indicating a better binding geometry for L-than for D-aspartate. For both substrates, the DS contribution was unfavorable ( Table 1). When plotting the observed K d values for L-and D-aspartate against the sodium concentration (on logarithmic scales), the slopes of both curves in the lower limit of the sodium concentration are close to À3, indicating that binding of both compounds is coupled to the binding of three sodium ions (Boudker et al., 2007;Lolkema and Slotboom, 2015) ( Figure 1B).
To test whether D-aspartate is a transported substrate, purified Glt Tk was reconstituted into proteoliposomes and uptake of [ 3 H]-D-aspartate was assayed. Glt Tk catalyzed transport of the radiolabeled substrate into the proteoliposomes. The K m for transport was 1.1 ± 0.11 mM at a sodium concentration of 100 mM ( Figure 1C). This value is comparable to the K m for L-aspartate uptake under the same conditions (0.75 ± 0.17 mM). The stoichiometry Na + : D-aspartate was determined by flux measurements of radiolabeled D-aspartate at different membrane voltages (Fitzgerald et al., 2017). Depending on the concentrations of Na + and D-aspartate on either side of the membrane, the imposed voltages either lead to flux of radiolabeled D-aspartate across the membrane (accumulation into or depletion from the lumen), or does not cause net flux (when the voltage equals the equilibrium potential) (Fitzgerald et al., 2017). The equilibrium potentials for different possible stoichiometries are calculated by: where n and m are the stoichiometric coefficients for Na + and substrate S, respectively. Membrane voltages were chosen that would match the equilibrium potential for stoichiometries of 2:1 (À78 mV), 3:1 (À39 mV) or 4:1 (À26 mV), and flux of radiolabeled D-aspartate was measured ( Figure 1D). At À78 mV D-aspartate was taken up into the lumen; at À26 mV it was released from the liposomes; and at À39 mV there was little flux. From these data, we conclude that D-aspartate is most likely symported with three sodium ions. However, the flux was not exactly zero at the calculated equilibrium potential of À39 mV for 3:1 stoichiometry. This small deviation could be caused by systematic experimental errors, or by leakage or slippage (Parker et al., 2014;Shlosman et al., 2018). To exclude that it was caused specifically by D-aspartate, we repeated the experiment using radiolabeled L-aspartate. The equilibrium potentials for the experiments using D-and L-aspartate were identical, showing that the two stereoisomers use the same coupling stoichiometry.

Similar mode of enantiomers binding
We determined a crystal structure of Glt Tk in complex with D-aspartate at 2.8 Å resolution (Figure 2A,B). The obtained structure is highly similar to the previously described Glt Tk and Glt Ph . Insets show no D-aspartate binding in absence of NaCl. (B) Sodium and aspartate binding stoichiometry. Logarithmic plot of K d values (nM) for L-aspartate (black squares; slope is À2.8 ± 0.4; taken for reference from Guskov et al., 2016) and D-aspartate (gray circles; slope is À2.9 ± 0.2) against logarithm of NaCl concentration (mM). The negative slope of the double logarithmic plot (red line) in the limit of low sodium concentrations indicates the number of sodium ions that bind together with aspartate. Error bars represent the ±SD from at least three independent measurements. (C) Glt Tk transport rate of D-aspartate in presence of 100 mM NaCl. The solid line reports the fit of the Michaelis-Menten model to the data revealing a K m value of 1.1 ± 0.11 mM. Error bars represent the ±SD from duplicate experiments. (D) Determination of Na + : aspartate coupling stoichiometry in Glt Tk using equilibrium potential measurement. The uptake or efflux of radiolabeled aspartate was determined by comparing the lumenal radioactivity associated with the liposomes after 2 min of incubation with the radioactivity initially present (Dcpm). Gray circles and black squares show the measurements for D-and L-aspartate, respectively. The solid and dashed lines are the best linear regression for the D-and L-aspartate data, Figure 1 continued on next page structures with the transport domains in the outward-oriented occluded state. Comparison of the Glt Tk structures in complex with L-and D-aspartate revealed a highly similar binding mode of the substrates with analogous orientation of amino and carboxyl groups. Despite the impossibility to superimpose two enantiomers, D-and L-aspartate are capable of forming almost identical hydrogen bonding networks with conserved amino acid residues of the substrate-binding site ( Figure 2C).
There are only small changes in the positions of the Ca atoms and Cb carboxyl groups due to the constitutional differences. However, this divergence leads to only minor changes in the interaction network, consistent with the comparable K d and DH values determined by ITC ( Table 1). Three peaks of electron density ( Figure 2D; Figure 2-figure supplement 2) located at the same positions as three sodium ions in the Glt Tk complex with L-aspartate (Guskov et al., 2016) most probably correspond to sodium ions, consistent with a 3:1 Na + : D-aspartate coupling stoichiometry ( Figure 1B,D).

Discussion
Most proteins selectively bind a single stereoisomer of their substrates (for a review see Nguyen et al., 2006). On the other hand, some proteins are able to bind different stereoisomers of a ligand, which is believed to be possible due to different binding modes, because enantiomers cannot be superimposed in the three-dimensional space and thus cannot interact with the binding site identically.
Based on three-and four-point attachment models (Easson and Stedman, 1933;Mesecar and Koshland, 2000;Ogston, 1948) it has been suggested that stereoisomers can bind in the same site but with significant differences. This hypothesis was supported by crystal structures of enzymes with different enantiomeric substrates (Brem et al., 2016;Sabini et al., 2008), including enantiomeric amino acids (Aghaiypour et al., 2001;Bharath et al., 2012;Driggers et al., 2016;Temperini et al., 2006). In contrast, the binding poses of enantiomers in some other enzymes are remarkably similar, for instance in aspartate/glutamate racemase EcL-DER, where active site forms pseudo-mirror symmetry (Liu et al., 2016).
To our knowledge Glt Tk is the first amino acid transporter for which the binding of enantiomeric substrates has been characterized. The only other transporter for which structures have been determined in the presence of D-and L-substrates is the sodium-alanine symporter AgcS. However, in that case, limited resolution prevented determination of the absolute orientation of bound enantiomers (Ma et al., 2019). In the substrate-binding site of Glt Tk , L-and D-aspartate take similar poses leading to almost identical networks of contacts. Since mirror imaged substrates inevitably have respectively. The 95% confidence interval for D-aspartate is displayed by gray curves. Numbers in parentheses are the coupling stoichiometries expected to give zero flux conditions for each membrane voltage. Error bars represent the ± SD obtained in five replicates. DOI: https://doi.org/10.7554/eLife.45286.002 The following source data is available for figure 1: Source data 1. Final concentrations of internal and external buffer used in each reversal potential experiment after diluting the proteoliposomes. DOI: https://doi.org/10.7554/eLife.45286.003 Table 1. Thermodynamic parameters of D-and L-aspartate binding at high (300 mM) and low (75 mM) Na + concentration.  differences in angles between donors and acceptors of hydrogen bonds, the binding affinities are not identical, with 4-6 times higher K d of the Glt Tk -D-aspartate complex in comparison with L-aspartate (Table 1). Similar differences in binding affinities between these enantiomers were also found for the Glt Ph homologue (Boudker et al., 2007). The higher K d values for the D-aspartate enantiomer might be explained by a higher dissociation rate (k off ) in comparison with L-aspartate, that was shown in kinetic studies of sodium and aspartate binding on Glt Ph (Ewers et al., 2013;Hänelt et al., 2015). Glt Tk couples binding and transport of three sodium ions to one D-aspartate molecule ( Figure 1B,D), the same number as for L-aspartate. Although the affinity for D-aspartate is lower than for L-aspartate, the binding of D-aspartate is not accompanied by a loss of sodium binding sites, which is in line with the observation that none of the sodium binding sites are directly coordinated by the substrate L-aspartate. In the crystal structure of Glt Tk with D-aspartate peaks of density were resolved at positions corresponding to the three sodium ions in the L-aspartate bound Glt Tk structure ( Figure 2D) (Guskov et al., 2016). Altogether our data suggest that the mechanism of D-and L-aspartate transport in Glt Tk is most probably identical.
Mammalian glutamate transporters take up D-aspartate, L-glutamate and L-aspartate with similar micromolar affinity, but have significantly lower affinity (millimolar) for D-glutamate (Arriza et al., 1997;Arriza et al., 1994). In the absence of the structures of human SLC1A transporters with different stereoisomeric substrates, one can only speculate why EAATs can readily bind and transport both L-and D-aspartate, but only L-glutamate. It seems that the extra methylene group in D-glutamate compared to D-aspartate could cause sterical clashes within the binding site (Figure 2 -Figure 2-figure supplement 3), which might affect affinity of binding.

Protein purification and crystallization
Glt Tk was expressed and purified as described previously (Guskov et al., 2016). It was shown that L-aspartate binds to Glt Tk only if sodium ions are present, and the protein purified in absence of sodium ions is in the apo state (Jensen et al., 2013). For crystallization with D-aspartate the apo protein was purified by size exclusion chromatography (SEC) on a Superdex 200 10/300 GL (GE Healthcare) column equilibrated with buffer containing 10 mM Hepes KOH, pH 8.0, 100 mM KCl, 0.15% DM. Crystals of Glt Tk with D-aspartate were obtained in presence of 300 mM NaCl, 300 mM D-aspartate (Sigma-Aldrich, 99%) by the vapour diffusion technique (hanging drop) at 5˚C by mixing equal volumes of protein (7 mg ml À1 ) and reservoir solution (20% glycerol, 10% PEG 4000, 100 mM Tris/bicine, pH 8.0, 60 mM CaCl 2 , 60 mM MgCl 2 , 0.75% n-octyl-b-D-glucopyranoside (OG)).

Data collection and structure determination
Crystals were flash-frozen without any additional cryo protection and data sets were collected at 100K at the beamline ID23-1 (ESRF, Grenoble). The data were indexed, integrated and scaled in XDS (Kabsch, 2010) and the structure was solved by Molecular Replacement with Phaser (McCoy et al., 2007) using structure of Glt Tk (PDB ID 5E9S) as a search model. Manual model rebuilding and refinement were carried out in COOT (Emsley et al., 2010) and Phenix refine (Afonine et al., 2012). Data collection and refinement statistics are summarized in Table 2. Coordinates and structure factors for Glt Tk have been deposited in the Protein Data Bank under accession codes PDB 6R7R. All structural figures were produced with an open-source version of PyMol.

Isothermal titration calorimetry
ITC experiments were performed at a constant temperature of 25˚C using an ITC200 calorimeter (MicroCal). Varying concentrations of the indicated substrates (in 10 mM Hepes KOH, pH 8.0, 100 mM KCl, 0.15% DM and indicated sodium concentrations) were titrated into a thermally equilibrated ITC cell filled with 250 ml of 3-20 mM Glt Tk supplemented with 0 to 1000 mM NaCl. Data were analyzed using the ORIGIN-based software provided by MicroCal.

Reconstitution into proteoliposomes
A solution of E. coli total lipid extract (20 mg ml À1 in 50 mM KPi, pH 7.0) was extruded with a 400nm-diameter polycarbonate filter (Avestin, 11 passages) and diluted with the same buffer to a final concentration of 4 mg ml À1 . The lipid mixture was destabilized with 10% Triton-X100. Purified Glt Tk and the destabilized lipids were mixed in a ratio of 1:1600 or 1:250 (protein: lipid) and incubated at room temperature for 30 min. Bio-beads were added four times (25 mg ml À1 , 15 mg ml À1 , 19 mg ml À1 , 4 mg ml À1 lipid solution) after 0.5 hr, 1 hr, overnight and 2 hr incubation, respectively, on a rocking platform at 4˚C. The Bio-beads were removed by passage over an empty Poly-Prep column (Bio-Rad). The proteoliposomes were collected by centrifugation (20 min, 298,906 g, 4˚C), subsequently resuspended in 50 mM KPi, pH 7.0 to the concentration of the protein 33.4 mg ml À1 and freeze-thawed for four cycles. The proteoliposomes were stored in liquid nitrogen until subsequent experiments.

Uptake assay
Stored proteoliposomes with reconstitution ratio of 1:1600 were thawed and collected by centrifugation (20 min, 298,906 g, 4˚C), the supernatant was discarded and the proteoliposomes were resuspended in buffer containing 10 mM KPi, pH 7.5, 300 mM KCl. The internal buffer was exchanged by three cycles of freezing in liquid nitrogen and thawing, and finally extruded through a polycarbonate filter with 400 nm pore size (Avestin, 11 passages). The proteoliposomes were finally pelleted by centrifugation (20 min, 298,906 g, 4˚C) and resuspended to the concentration of the protein 625 ng ml À1 . 2 ml of proteoliposomes were diluted 100 times in reaction buffer containing 10 mM KPi, pH 7.5, 100 mM NaCl, 200 mM Choline-Cl, 3 mM valinomycin and 0.2-15 mM D-aspartate (each concentration point contained 0.2 mM [ 3 H]-D-aspartate). After 15 s the reaction was quenched by adding 2 ml of ice-cold buffer (10 mM KPi, pH 7.5, 300 mM KCl) and immediately filtered on nitrocellulose filter (Protran BA 85-Whatman filter), finally the filter was washed with 2 ml of quenching buffer. The filters were dissolved in scintillation cocktail and the radioactivity was measured with a PerkinElmer Tri-Carb 2800RT liquid scintillation counter.

Measuring transporter equilibrium potentials
Stored proteoliposomes with reconstitution ratio of 1:250 were thawed and collected by centrifugation (20 min, 298,906 g, 4˚C), the supernatant was discarded and the proteoliposomes were resuspended to a concentration of 10 mg ml À1 of lipids in buffer containing 20 mM Hepes/Tris, pH 7.5, 200 mM NaCl, 50 mM KCl, 10 mM D-aspartate (containing 1 mM [ 3 H]-D-aspartate). The internal buffer was exchanged by freeze-thawing and extrusion as described above. The experiment was started by diluting the proteoliposomes 20 times into a buffer containing 20 mM Hepes/Tris, pH 7.5, 200 mM NaCl, 3 mM valinomycin, varying concentrations of KCl and Choline Cl were added in order to obtain the desired membrane potential as shown in (Figure 1-source data 1). After 1, 2 and 3 min the reaction was quenched with ice-cold quenching buffer containing 20 mM Hepes/Tris, pH 7.5, 250 mM Choline Cl and immediately filtered on nitrocellulose filter (Protran BA 85-Whatman filter), finally the filter was washed with 2 ml of quenching buffer. The initial amount of radiolabeled aspartate was measured by filtering the proteoliposomes immediately after diluting them in quenching buffer. The filters were dissolved in scintillation cocktail and the radioactivity was measured with a PerkinElmer Tri-Carb 2800RT liquid scintillation counter. The equilibrium, or reversal, potential, E rev , for each condition was calculated as described in Fitzgerald et al. (2017). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.