Glutamine transport across cell membranes is performed by a variety of transporters, including the alanine serine cysteine transporter 2 (ASCT2). The substrate-binding site of ASCT2 was proposed to be specific for small amino acids with neutral side chains, excluding basic substrates such as lysine. A series of competitive inhibitors of ASCT2 with low µM affinity were developed previously, on the basis of the 2,4-diaminobutyric acid (DAB) scaffold with a potential positive charge in the side chain. Therefore, we tested whether basic amino acids with side chains shorter than lysine can interact with the ASCT2 binding site. Molecular docking of L-1,3-diaminopropionic acid (L-DAP) and L-DAB suggested that these compounds bind to ASCT2. Consistent with this prediction, L-DAP and L-DAB, but not ornithine, lysine or D-DAP, elicited currents when applied to ASCT2-expressing cells. The currents were carried by anions and showed the hallmark properties of ASCT2 currents induced by transported substrates. The L-DAP response could be eliminated by a competitive ASCT2 inhibitor, suggesting that binding occurs at the substrate binding site. The KM for L-DAP was weakly voltage dependent. Furthermore, the pH dependence of the L-DAP response showed that the compound can bind in several protonation states. Together, these results suggest that the ASCT2 binding site is able to recognize L-amino acids with short, basic side chains, such as the L-DAP derivative β-N-methylamino-l-Alanine (BMAA), a well-studied neurotoxin. Our results expand the substrate specificity of ASCT2 to include amino acid substrates with positively charged side chains.
The alanine serine cysteine transporter 2 (ASCT2, SLC1A5) is a neutral amino acid transporter that moves amino acids with uncharged side chains into cells, in exchange for intracellular amino acids [1–4]. ASCT2 belongs to the solute carrier 1 (SLC1) family of transport proteins that consists of two members that are specific for neutral amino acid substrates (ASCT1, SLC1A4  and ASCT2, SLC1A5 [6,7]) and five high affinity excitatory amino acid transporters 1–5 (SLC1A1–3 and SLC1A6–7) , which recognize acidic amino acid substrates. The subtype ASCT2 has been shown to transport amino acids with small hydrophilic side chains, such as serine, cysteine, asparagine, glutamine, and alanine, which have a shorter hydrophobic side chain [1,7,9,10].
ASCT2 has recently received significant attention after it was proposed to be actively involved in the growth of cancer cells, by satisfying specific metabolic and signaling needs [11–15]. ASCT2 is up-regulated in several cancer types, such as glioma , hepatoma , melanoma , neuroblastoma , colon , breast [20,21], prostate  and lung cancer [19,23], due to high energy demands of rapidly growing tissue that may not be adequately met by biosynthesis alone. These studies suggest that ASCT2 is an important pharmacological drug target . However, there is limited information on the pharmacology of ASCT2 and small molecule inhibitor binding. The newly solved crystal structures of EAAT1 in outward facing conformations , as well as cryo-EM structures of ASCT2 in inward-facing conformations can aid in understanding substrate binding [26,27]. In addition, various mutations of the archaean SLC1 glutamate transporter homolog GltPh (prokaryotic aspartate transporter from Pyrococcus horikoshii) have been made that convert the substrate specificity to recognition of neutral, rather than acidic amino acids . GltPh shares ∼23% sequence identity with ASCT2 [29,30]. These studies provide an important understanding of the molecular basis for substrate selectivity in the SLC1 family, which is crucial for the development of high affinity and selective ASCT2 inhibitors [25,28,29,31–33].
Recently, a group of amino acid analogs were shown to inhibit ASCT2 with low micromolar affinities . These compounds are based on 2-amino-4-bis(aryloxybenzyl) aminobutanoic acid (most potent inhibitor shown in Figure 1A). Under physiological conditions, these compounds may be protonated, thus bearing a positive charge at the side chain nitrogen. However, the parent compound 2,4-diaminobutanoic acid (L-DAB, Figure 1B), which has a basic side chain nitrogen, has not been tested for activity with respect to ASCT2. Lysine, another amino acid with basic side chain, failed to inhibit alanine uptake by ASCT2 [1,3,6]. Thus, it was concluded that native basic amino acids are not recognized by the ASCT2 binding site.
Structures of representative known ASCT2 transported substrates and inhibitors.
Another amino acid analog with basic side chain is β-N-Methylamino-l-Alanine (BMAA, Figure 1C). BMAA is a L-alanine derivative with a methylamino substitution in the side chain. BMAA is produced by cyanobacteria , and has been identified as a neurotoxin and a potential cause of neurodegenerative disease, for example Amyotrophic Lateral Sclerosis (ALS) . BMAA transport was shown to be blocked by the ASCT2 inhibitor benzylserine . Thus, it is possible that ASCT2 contributes to the patho-physiological effects of BMAA .
To test whether ASCT2 can recognize amino acid derivatives with basic side chains, we determined activity of ASCT2 in response to application of several amino acid analogs with varying side chain lengths. Interestingly, our results show that ASCT2 binds, and possibly transports small amino acids with an amino group in the side chain, for example L-2,3-diaminopropionic acid (L-DAP). However, consistent with previous results, ASCT2 does not bind or transport lysine [1,3]. Additionally, we found that apparent affinity increased with decreasing side chain length and L-DAP binding could be inhibited competitively using previously published ASCT2 inhibitors . We propose a mechanism for interaction of ASCT2 with various protonation states of L-DAP (Figure 1D–G), based on kinetic modeling as well as molecular docking. Overall, these findings extend the substrate specificity of ASCT2 to include small amino acid substrates with basic side chains.
Materials and methods
Rat ASCT2 (kindly provided by S. Bröer) and human ASCT2 (Origene) and YFP complementary DNAs were used to transiently co-transfect sub-confluent human embryonic kidney (HEK 293) cells using POLYPLUS Jet-prime transfection reagent according to the instructions of the supplier. Cell analyses were performed 24–30 h after transfection using electrophysiological techniques. Tested compounds were purchased from Alfa Aesar and BAC unless otherwise stated.
Electrophysiological experiments were performed as described previously in detail [31,39]. Compounds were dissolved in external buffer with concentrations ranging from in mM 0.05 to 5.00. External buffer consisted of in mM 140 NaCl, 2 CaCl2, 2 MgCl2 and 10 HEPES at pH 7.40 (NaOH). In brief, HEK293 cells immersed in the external buffer and expressing rat/human ASCT2 were suspended from a current recording electrode hosted by a micro glass pipette in whole cell configuration. Then compounds were consequently applied to cells via rapid solution exchange device as described in detail previously . Internal solution in the glass pipette consisted in mM of 130 NaSCN, 2 MgCl2, 10 EGTA, and 10 HEPES, pH 7.40 (NaOH). In voltage jump experiments, external and internal buffers were replaced with 140 mM NaMes and 130 mM NaMes, respectively. Series resistance of the open pipette was between 2.9–5.0 MΩ. Series resistance was not compensated due to relatively small currents. Currents traces were recorded using an Adams and List EPC7 amplifier and digitized using a Molecular Devices Digidata A/D converter.
As described previously , non-transfected cells and cells transfected with vector alone showed no significant response to application of alanine and other ASCT2 substrates.
Glutamine uptake fluorescence assay
Uptake studies using 14C BMAA were performed by BioIVT (Optiva Bioctechnology Inc, Santa Clara, CA). In brief, uptake was performed in ASCT2-transfected MDCK-II cells grown as monolayers in 96-well transwell membrane plates. Transport assays were carried out 48 h after transfection. An amount of 2 mM 14C BMAA was added to the apical side of the membrane, after washing. After 5 min incubation, solutions were aspirated at both apical and basolateral side, and both sides were washed. Cells were extracted using acetonitrile:water (50 : 50) and 30 µl of the extraction mix was added to 200 µl of scintillation fluid. Radioactivity was assayed using a PerkinElmer 1450 Microbeta scintillation counter.
Microcal Origin software was used to fit both linear and nonlinear curves. Standard linear equation (y = a + bx) was used in fitting linear plots. Pearson's r and adjusted R2 were then obtained from the fitted plot. Non-linear dose response relationships were fitted using Michaelis–Menten-like equation. KM and Imax (current at saturating substrate concentrations) were then obtained from the fitted data. Experiments were performed at least four times and at least four cells per experiment. Error bars in all graphs represent mean ± SD.
While cryo-EM structures of ASCT2 have been solved in inward facing conformations, the goal of this study is to characterize ligands likely encountering the substrate binding site from the extracellular space. Thus, we focused on docking the ligands to the binding site of our previously published models of ASCT2 based on the outward-occluded and outward-open structures of the human EAAT1 transporter . Docking was performed using Glide from the Schrödinger suite . The models were prepared for docking with the Maestro Protein Preparation Wizard under default parameters. The binding site was defined using the co-ordinates of a reference ligand from the template structure, EAAT1 (aspartate or (3S)-3-[[3-[[4-(Trifluoromethyl)benzoyl] amino]phenyl]methoxy]-l-aspartic acid, TFB-TBOA) with the Maestro Receptor Grid Generation panel. Constraints were used for docking and varied by conformation. For the outward-occluded model, we used hydrogen bond constraints on S351 and N471. For the outward-open model, we used hydrogen bond constraints at S351 or S353 on hairpin 1 (HP1), as previously described . We also removed UCPH101 prior to assigning the grid for the outward-open model to avoid multiple reference ligands in the box. Small molecules were prepared for docking using LigPrep of the Schrödinger suite . Default settings were used except for the ‘Ionization’ parameter, which was set to ‘Do not ionize’ because ionization states were manually specified per molecule. PyMOL was used to visualize the docking results .
Computational analysis of basic amino acid interaction with ASCT2 using docking
Here, we aimed to investigate ASCT2 amino acid substrate specificity using small molecules with basic amino acid side chains. We used L-2,3-diaminopropionic acid (L-DAP) as the molecule with the shortest basic side chain, docking it to our ASCT2 homology model generated on the basis of the EAAT1 crystal structure in both the outward-occluded (substrate-bound) and outward-open (inhibitor-bound, HP2 open) conformations. Docking was performed in four possible protonation states of L-DAP (Figure 1D–G). For the outward-occluded state, we observed interactions between the α-carboxy group of the ligand and the N471 nitrogen atom and S353 backbone nitrogen, as well as the α-amino group of L-DAP with S351 (main chain carbonyl) and the D464 side chain (Figure 2). Interactions were also observed with the side chain of T468. These interactions were also found for known amino acid substrates of ASCT2 without basic side chain, for example L-serine and L-asparagine.
Docking poses of L-DAP and L-DAB.
It should be noted that the neutral, zwitterionic form of L-DAP with the α-amino group deprotonated is the major predicted protonation state at physiological pH (Figure 1F), as determined by Hay and Morris . Additional information about protonation states of L-DAP is provided in the Discussion section. Specifically, in the outward-occluded state, the neutral L-DAP side chain amino group is predicted to be co-ordinated by the backbone oxygen atoms of P432 and I431 in HP2 (Figure 2A,B). Similar interactions were observed for L-2,4-diaminobutyric acid (L-DAB), which has an additional CH2 group in the side chain compared with L-DAP that is predicted to form additional polar contacts with D464 (Figure 2C).
The compounds were also docked to the outward-open state, in which the tip of HP2 moves away from the binding site, opening it up to the extracellular side of the membrane. Docking scores were generally lower in this configuration, presumably because of the lack of polar interaction of the side chain amino group with residues in HP2 (Supplementary Figure S1). In both, outward-occluded and outward-open conformations, docking scores did not vary substantially between the four protonation states that were tested. The neutral L-DAP scored better than other forms in the outward-open state (Supplementary Table S1).
Amino acids with basic side chains show substrate-like properties
To test the predictions from computational analysis, we performed patch clamp recordings of HEK293 cells overexpressing ASCT2 upon application of amino acids with basic side chains. Interestingly, at pH 7.40 and at 2 mM concentration, L-DAP elicited an inwardly directed current very similar to the one produced when alanine, an ASCT2 substrate, is applied [1,3,31,44], as shown in Figure 3A,B. SCN− at 140 mM concentration was used in the intracellular solution, which flows out of the cell upon substrate application, due to activation of the anion conductance [2,39]. This anion outflow creates inward current [1,3,31,44]. We also used 10 mM alanine and 130 mM Na+ in the internal solutions, to push the transporter to the outward-facing conformations (i.e. outward-occluded or outward-open) best suited for analyzing substrates and inhibitors. In contrast with L-DAP, D-2,3-DAP did not produce any current up to a tested concentration of 4 mM, suggesting ASCT2 specificity for L amino acids with basic side chains. L-DAP induced currents increased with increasing concentration and were close to saturation at ∼2 mM concentration. The dose-response relationship of this current could be fitted using a Michaelis–Menten-like equation (see Materials and Methods) to obtain the apparent affinity of L-DAP of 560 ± 70 µM, as shown in Figure 3C. L-2,4-diaminobutyric acid (L-DAB) also produced an inward current, however, with smaller magnitude than that generated by L-DAP at the same concentration (Figure 3B). The L-DAB dose-response relationship curve could not be evaluated, due to much lower affinity as compared with L-DAP (Figure 3C). Since no saturation was observed at concentrations up to 5 mM, we were unable to obtain the apparent affinity. However, this affinity is clearly much lower than that for L-DAP.
Basic amino acids generate current responses in ASCT2.
We extended our scope to molecules with even longer basic side chains. The inwardly directed current decreased with increasing side chain length with L-ornithine and L-lysine producing very small currents at 4 mM that were not significantly different from zero, as shown in Figure 3A. These results are consistent with previous observations showing that L-lysine is not a transported substrate of ASCT2 [1–3]. Finally, homoglutamine, which has an additional carbon in the side chain compared with glutamine, showed inhibitor-like properties (outward current, Figure 3A), suggesting that amino acids with increased chain length, but neutral side chain can bind, but cannot be transported by ASCT2.
L-DAP competes with binding of substrate and a competitive inhibitor
To investigate further if the compounds with basic side chains were actually binding to the ASCT2 binding site, we conducted competition experiments at varying concentrations of a competitive inhibitor of ASCT2, (R)-γ-(4-biphenylmethyl)-l-proline (BPP). This compound was previously characterized and competes with alanine for the substrate binding site with a Ki of 3 µM . Consistent with the hypothesis of L-DAP acting as a substrate, BPP reduced L-DAP induced inward currents at all concentrations and the results could be fitted with a Michaelis–Menten-like relationship to obtain the apparent binding affinity for L-DAP at three different inhibitor concentrations (Figure 4A–C). Figure 4D shows the KM for L-DAP plotted vs. the inhibitor concentration and as expected for competitive inhibition, a linear relationship was obtained with a Pearson's r value of −0.992 and R2 as 0.970 demonstrating very good correlation. These results demonstrate that L-DAP is binding to the same site as the inhibitor, further supporting the idea that DAP is a substrate.
L-DAP binds to the substrate binding site.
We also tested whether L-DAP can block substrate flux (Figure 4E). To test this possibility, we utilized a fluorescent glutamine sensor as an assay for glutamine influx. HEK293 cells transiently transfected with rASCT2 and the sensor FLIPQ-TV3.0_100 µM (kindly provided by S. Okumoto, ) demonstrate that L-DAP and benzylserine block glutamine flux into cells via rASCT2 transport. Cells showed a 11.2 ± 0.89% decrease in fluorescence emission in the Venus range when 500 µM glutamine was applied externally, indicating glutamine uptake. The same cells showed only a 1.00 ± 1.14% decrease in fluorescence emission in the Venus range when 500 µM glutamine plus 2 mM L-DAP were applied, demonstrating that L-DAP blocks glutamine uptake by rASCT2. For comparison, 500 µM glutamine plus 5 mM benzylserine  produced a 2.70 ± 1.25% decrease in fluorescence emission in the Venus range.
L-DAP-induced current is carried by anions
Previous studies have shown that the substrate-induced ASCT2 current is carried by anions [31,38,39,47]. To test the ionic basis of the current induced by L-DAP, we tested its voltage dependence. First, 140 mM NaSCN was used as an extracellular buffer and 130 mM NaSCN on the intracellular side. Under these conditions, the reversal potential of the current, if carried by anions, should be very close to 0 mV (−1.9 mV according to the Goldman equation) . Consistent with this expectation, L-DAP and L-DAB-induced currents were inward directed at negative potentials, and outward directed at positive potentials, as shown in Figure 5A. When we performed a similar experiment, but replaced the extracellular buffer SCN− with 140 mM Cl−, an anion with low permeability, currents evoked by rapid application of alanine, L-DAP, and L-DAB were inwardly directed at all voltages, but increased with increasingly negative membrane potential (Figure 5B). Finally, L-DAP failed to elicit currents in the absence of permeating anions (replacement by methane sulfonate, data not shown). Together, these results are in agreement with the interpretation that L-DAP and L-DAB induced currents are caused by anion conductance that is activated upon binding of these compounds, in analogy to known transported substrates, such as alanine [1,2,47].
Current induced by basic amino acids is carried by anions.
It should be noted that the anion current carried by SCN− is not a physiological phenomenon. While the anion current may be present in chloride-containing solutions (the major physiological anion), its magnitude is too small to test substrate/inhibitor activity. In fact, the anion selectivity sequence closely matches that found in the related glutamate transporters of the SLC1 family, with P(SCN−) >> P(NO3−) > P(I−) > P(Cl−), as shown by Bröer et al. . Unfortunately, due to the electroneutral nature of ASCT2 amino acid exchange, the anion conductance is the only possibility to study ASCT2 using electrophysiology.
The apparent affinity of ASCT2 for L-DAP, but not alanine is voltage dependent
To further investigate the role of the basic side chain of L-DAP, we analyzed apparent affinities of L-DAP at different cell membrane potentials, in comparison with the neutral substrate alanine. As expected, there was almost no change of apparent alanine affinity with voltage (Figure 6, circles) because, alanine is a zwitterion and is uncharged at pH 7.4. Conversely, L-DAP showed a slight increase in apparent affinity from positive to negative membrane potentials, suggesting that L-DAP, with a net positive charge, binds more strongly at negative potential (Figure 6, squares). The apparent valence calculated from the slope of the log (KM) vs. voltage relationship is 0.21 ± 0.04. Therefore, it is possible that L-DAP senses only a fraction of the membrane electric field while binding to the ASCT2 binding site, or that, at pH 7.4, only a fraction of L-DAP binds in the positively charged form, or that both of these scenarios apply in combination.
L-DAP binding senses the transmembrane electric field.
Effect of protonation state of L-DAP on ASCT2 interaction
One possibility is that L-DAP binds to ASCT2 in the state with a deprotonated side chain, even though this is only a small fraction of molecules at pH 7.4. To test this possibility, we determined L-DAP currents and apparent affinities as a function of extracellular pH. In all experiments, we kept bath and intracellular solutions at pH 7.4 to maintain cell viability, only switching to pH values different from 7.4 for short periods of time (10 s). A concentration of 1 mM alanine was selected as a control in all experiments because it was shown in the literature to exhibit a KM that is almost independent of the pH . L-DAP induced currents increased with pH, with the largest currents observed at pH 10.4, as shown in Figure 7A. At this pH, L-DAP is primarily present in the deprotonated form (Figure 1D). Surprisingly at pH below pH 6.50, currents evoked by application of 2 mM L-DAP were outwardly directed (Figure 7A). At this pH, L-DAP is primarily present in the state that is protonated at both amino groups, as shown in Figure 1G. This behavior shows that doubly protonated L-DAP interacts with ASCT2, possibly inhibiting the leak anion current. This behavior is the hallmark of ASCT2 competitive inhibitors [31,38,39,44]. L-DAP apparent KM as a function of the pH showed a linear relationship (Figure 7C), with the KM increasing with decreasing proton concentration. These results suggest that the protonated form of L-DAP, while behaving as an inhibitor (outward current), binds more strongly to ASCT2 than the deprotonated form, which, on the other hand, behaves analogously to a transported substrate (inward current). Motivated by these results, we propose a mechanism for interaction of ASCT2 with various major protonation states of DAP shown in Figure 1D–G. This mechanism allowed us to predict the current/KM vs. pH relationships and is detailed further in the Discussion section. The predicted L-DAP maximum current response as a function of the pH agrees very well with experimental data, Figure 7B. The calculated KM vs. pH relationship showed linear increase in KM with pH, in agreement with experimental results. These results demonstrate that L-DAP binds more strongly at lower pH, while possibly becoming an inhibitor in the fully protonated state.
L-DAP can interact with ASCT2 in several protonation states.
The neurotoxin β-N-Methylamino-L-Alanine (BMAA) activates ASCT2 anion current
We next tested the L-DAP analog BMAA with ASCT2-expressing HEK293 cells. Like L-DAP, BMAA application elicited inward currents, but of significantly smaller magnitude than L-DAP and alanine (less than 20% of the maximum anion current induced by L-alanine, Figure 8A,B). The currents were dose dependent, and almost saturating at highest BMAA concentration used, 2 mM with a KM value of 1400 ± 320 µM (Figure 8C). We also tested uptake of 14C BMAA, however, while the uptake at 2 mM concentration (193 ± 24 pmol/min/cm2), was slightly higher in ASCT2-expressing cells compared with control cells (163 ± 36 pmol/min/cm2), this increase was not significant. Most likely, BMAA uptake, if it occurs, is too slow to be detected within the sensitivity of the uptake assay. These results suggest that BMAA shows weak substrate-like properties with respect to ASCT2 anion conductance activation, but may not be a transported substrate, or is transported by ASCT2 at a rate at least 10-fold lower than L-glutamine.
L-BMAA shows substrate-like properties.
The glutamine transporter ASCT2 has become increasingly important as a potential cancer drug target. Several studies have shown that targeting by siRNA [13,21,49–51] and small molecule inhibitors [24,52], limits cell growth in rapidly growing tumor tissue. However, there are still substantial gaps with respect to our understanding of substrate and inhibitor specificity and mechanisms of interaction with ASCT2. It has been generally accepted that ASCT2 is a neutral amino acid transporter that cannot transport positively charged amino acids [1,3]. The major findings from our present work are: (1) amino acids with basic side chains can bind to the ASCT2 substrate binding site, showing properties of transported substrates at physiological pH; (2) the potency of interaction decreases with the length of the basic side chain (ornithine and lysine are excluded from the binding site); (3) the affinity of interaction and ability to induce a current response depends on pH and, thus, protonation state of the basic amino acid, in contrast with interaction of the neutral substrate alanine; and (4) the neurotoxin BMAA, which has been implicated with the development of several neurodegenerative diseases , binds to ASCT2 and shows properties of a transported substrate [1,2,31,39]. Together, these findings demonstrate that the scope of substrate specificity of ASCTs must be expanded to include amino acids with charged, basic side chains. Interestingly, other amino acid transporters, which were thought to be selective for neutral amino acids, such as LAT1, have been found to transport amino acid substrates with ionizable side chains . Thus, detailed studies of SLC substrate and inhibitor selectivity may uncover previously unknown functional similarities/differences between SLC families and render transporters more promiscuous than previously believed.
Consistent with results previously obtained for ASCT2 substrates, the current evoked by application of basic amino acids is carried by anions. While transported substrates activate the ASCT2 anion conductance, non-transportable inhibitors block the anion conductance [2,31,47]. In agreement with this, no measurable currents were generated in the absence of permeating anions, suggesting that extracellular amino acid exchange with intracellular alanine is either electroneutral, or too slow to generate current above the detection limit of electrophysiology (methane sulfonate was used instead of Cl− and SCN− (data not shown)). However, the apparent KM for L-DAP was found to be weakly dependent on the membrane potential, in contrast with the KM for the neutral amino acid substrate alanine , which was independent of the voltage. We speculate that the amino acid binding site is located partially within the transmembrane electric field, leading to attraction of a positively charged substrate to the binding site by an inside-negative membrane potential. Similar conclusions were reached for the binding of negatively charged amino acids and inhibitors to the acidic amino acid transporters (EAATs) sharing the same solute carrier family with the ASCTs . Thus, it is not unexpected that the binding mechanisms of substrates to the extracellular facing binding site are conserved within the family members.
Additional information about the mechanism of basic amino acid interaction was obtained by characterizing the pH dependence of KM and maximum substrate-induced current, Imax. In previous studies, it was shown that these parameters are largely pH-independent for neutral amino acid substrates, such as alanine, in a pH range in which the neutral zwitterion is the dominant species , suggesting that pH-dependent protonation changes of the transporter itself are not responsible for the pH effects on L-DAP KM and Imax. In contrast with neutral substrates like alanine, the KM and Imax for L-DAP were strongly dependent on the pH, indicating that L-DAP protonation state, rather than ASCT2 protonation state of critical residues, causes the functional changes in the pH range from 6.1 to 10.5. Furthermore, varying protonation states of L-DAP interact with ASCT2 with somewhat varying affinities and abilities to induce anion current. L-DAP is an unusual amino acid, because the pKa value of the α-amino group is much lower, 6.7,  than for other amino acids (compare to 9.1 for lysine, for example), meaning, this group is only protonated to 20% at pH 7.4, in contrast with 98% protonation of the α-amino group in lysine. Therefore, L-DAP is expected to exist as a mixture of predominately neutral protonation state (positive charge on the side-chain nitrogen, neutral in the α-position, Figure 1F, AH in eqn. 1), but partly in the fully protonated, positively charged form (Figure 1G, AH2 in eqn. 1). The neutral protonation state with a positive charge in the α-amino position, but no charge on the β-nitrogen (Figure 1E, pKa of 9.6), however, is expected to be present at small concentrations at all pH values tested.
We were able to describe the relationships between KM/Imax and pH with a simplified pre-equilibrium model (eqns. 1–4, appendix), taking into account three major protonation states of L-DAP, A, AH and AH2 (A = L-DAP, eqn.1, appendix).We assume that each protonation state of L-DAP interacts with the ASCT2 substrate binding site with a defined dissociation constant, K1–3 (eqns. 2–4, K1 = 2 mM, K2 = K3 = 200 µM), resulting in L-DAP bound states, TA, TAH, and TAH2 (eqns. 2–4, appendix). It was further assumed that each of these substrate-bound states catalyzes anion conductance with varying efficiencies (relative conductances listed in the figure legend of Figure 7), while the substrate-free state, T, cannot conduct anions. These assumptions are justified by previous reports on the mechanism of the anion conductance of ASCT2  as well the related glutamate transporters , showing that transport, i.e. movement of substrate across the membrane, is not necessary for anion current to be activated. This pre-equilibrium approach enabled us to model the populations of the transporter bound to A, AH and AH2 as a function of the pH, reproducing the experimental findings for the pH dependence of the KM for L-DAP (Figure 7B), as well as that of the anion current (Figure 7D). Irrespective of the exact distribution of these three protonation states, the model allows the following conclusions: (1) The fully deprotonated, negatively charged state, A, interacts with ASCT2 with low affinity, but with substrate-like behavior (state TA associated with anion current). (2) The fully protonated, positively charged state, AH2, interacts with ASCT2 with higher affinity (200 µM), but displays hallmark properties of an inhibitor, i.e. blockage of the leak anion current [31,39,44]. (3) The intermediate, neutral protonation state, AH, binds also with relatively high affinity (200 µM) and activates anion conductance, but less efficiently than the A state. Thus, AH has substrate-like behavior.
We recently published ASCT2 homology models based on the EAAT1 structure in both outward-occluded and outward-open conformations [42,44]. In the outward-open state, HP2 shows a more pronounced opening, generating a larger binding site suitable for the binding of inhibitors. It is assumed that inhibitor binding locks the transporter in the outward-open conformation, which is unable to undergo conformational changes needed for substrate transport . While crystal structures of ASCT2 in substrate-bound and substrate-free conformations are known, they are in the inward-facing conformation and, thus, not suitable for docking potential binders that act from the extracellular side [26,56]. It should be noted that while this work was in progress, an outward-occluded crystal structure with bound glutamine was published . This structure confirms the molecular interactions with substrates in the binding site predicted from our EAAT1 homology model [42,44], but is not available for docking analysis. The docking results with L-DAP in four protonation states in both outward-occluded and outward-open states show similar coordination of the α-carboxy and amino groups and confirm the general findings from the experiment data, namely that L-DAP can interact with ASCT2, and the three relevant protonation states interact with the binding site with docking scores in the same range.
It is noteworthy that in the outward-open state , a L-DAP configuration is predicted, in which the positions of the α and β-amino groups are switched, resembling a β-amino acid binding to the ASCT2 binding site. Although, we have found no evidence in the literature for the effect of β-amino acids on ASCT2, we speculate that it is this β-configuration that gives rise to the inhibitory effect of the fully protonated species at low pH. Second, our docking, as well as experimental results indicate that the α-amino group does not necessarily need to be protonated for interaction with ASCT2, especially if there is polar interaction in the side chain. This finding may be important, possibly allowing us in the future to extend the pharmacology of ASCT2 inhibitors by altering the α-NH2 function.
In addition to the structure-function implications of ASCT2 substrate specificity, our results have implications on the interaction of the non-proteinogenic amino acid BMAA with ASCT2 in the central nervous system. BMAA is a derivative of L-DAP, differing by only a methyl group on the β-amino group. BMAA, and to a lesser extent L-DAB, have been characterized as potent neurotoxins, potentially acting on neurotransmitter receptors in the mammalian brain. Both compounds are found in samples of cyanobacteria , which during algae blooms can contaminate food sources and move through the food chain into food products. Transport of BMAA and L-DAB by ASCT2 would be a compelling mechanism for the appearance of these compounds in the brain, as well as their transport into cells and eventual incorporation into proteins. Although the involvement of ASCT2 in blood brain barrier (BBB) transport of amino acids is less than clear, one report provides evidence for its expression in capillary endothelial cells, a model system of the BBB . Therefore, it is possible that BMAA gets access to the cerebrospinal fluid through ASCT2, in exchange for other amino acids, although other transport pathways at the BBB cannot be excluded. More direct evidence has been obtained from mammary gland cells, which allow movement of neurotoxic BMAA from the mother's blood stream to the milk during lactation, and, thus transfer to the off-spring  it was shown that benzylserine  inhibits BMAA transport in a mammary cell line, pointing to involvement of ASCT2. While our results are consistent with these findings, uptake of 14C BMAA by ASCT2 was not significant over control, non-ASCT2 expressing cells. Therefore, it is likely that uptake, if it occurs, is at a rate at least 10-fold lower than that of transported substrates, such as glutamine.
In conclusion, the work presented here provides direct evidence that amino acids with basic side chains can interact with the neutral amino acid transporter ASCT2, expanding the scope of molecular interactions in the substrate binding site. Like the transported substrate alanine, these basic amino acids elicit anion current, but in contrast with alanine bind in a voltage dependent manner. In addition to model compounds, the neurotoxin BMAA was found to be an activator of ASCT2 anion conductance, although uptake could not be ascertained within the sensitivity of the uptake assay. Further work will be required to study BMAA distribution in organisms and how it is affected by blocking ASCT2.
The authors declare that there are no competing interests associated with the manuscript.
This study was supported by a grant from the National Institutes of Health (http://www.nih.gov) (R01 GM108911) to A.S. and C.G. and T32 CA078207 to R.A.G., and by the National Science Foundation Grant 1515028 awarded to C.G.
E.N. performed the experiments and wrote the manuscript draft. R.A.G. and A.S. performed molecular modeling and docking. L.Z. performed the fluorescence glutamine uptake studies, analyzed and interpreted the results. E.N., R.A.G., A.S. and C.G. interpreted the data and compared modeling and experimental results. C.G. and A.S. designed the research. E.N., R.A.G., A.S. and C.G. contributed to the writing of the final manuscript.
The authors are grateful to Dr. Stefan Bröer for providing rat ASCT2 cDNA, and to Dr. Sakiko Okumoto for providing the fluorescent glutamine sensor FLIPQ-TV3.0_100 µM.
amyotrophic lateral sclerosis
alanine serine cysteine transporter 1
alanine serine cysteine transporter 2
blood brain barrier
excitatory amino acid transporter 1
ethylene glycol-bis(β-aminoethyl ether)-N,N,N′,N′-tetraacetic acid
human embryonic kidney
large amino acid transporter 1
solute carrier 1
trifluoromethyl)benzoyl] amino]phenyl]methoxy]-l-aspartic acid
yellow fluorescent protein