Electrogenic Steps Associated with Substrate Binding to the Neuronal Glutamate Transporter EAAC1*

Glutamate transporters actively take up glutamate into the cell, driven by the co-transport of sodium ions down their transmembrane concentration gradient. It was proposed that glutamate binds to its binding site and is subsequently transported across the membrane in the negatively charged form. With the glutamate binding site being located partially within the membrane domain, the possibility has to be considered that glutamate binding is dependent on the transmembrane potential and, thus, is electrogenic. Experiments presented in this report test this possibility. Rapid application of glutamate to the wild-type glutamate transporter subtype EAAC1 (excitatory amino acid carrier 1) through photo-release from caged glutamate generated a transient inward current, as expected for the electrogenic inward movement of co-transported Na+. In contrast, glutamate application to a transporter with the mutation A334E induced transient outward current, consistent with movement of negatively charged glutamate into its binding site within the dielectric of the membrane. These results are in agreement with electrostatic calculations, predicting a valence for glutamate binding of −0.27. Control experiments further validate and rule out other possible explanations for the transient outward current. Electrogenic glutamate binding can be isolated in the mutant glutamate transporter because reactions, such as glutamate translocation and/or Na+ binding to the glutamate-bound state, are inhibited by the A334E substitution. Electrogenic glutamate binding has to be considered together with other voltage-dependent partial reactions to cooperatively determine the voltage dependence of steady-state glutamate uptake and glutamate buffering at the synapse.

Plasma membrane glutamate transporters of the SLC1 (solute carrier 1) family actively take up glutamate from the extracellular space (1,2). The uptake process is energetically driven by the co-transport of three Na ϩ ions and one proton as well as the countertransport of one K ϩ ion (3,4). Binding of at least two of these Na ϩ ions has been shown to be dependent on the transmembrane potential, indicating that the cation binding sites are located at least somewhat within the membrane dielectric (5,6).
This idea is consistent with the available structural and computational data, showing that the Na1 binding site as well as the predicted Na3Ј sites are located close to the midpoint of the membrane (7,8). Electrogenic Na ϩ binding appears to be a hallmark of many Na ϩ -driven, secondary active transporters and has been also demonstrated for glucose transporters (9), GABA transporters of the SLC6 family (10), amino acid transporters (11), and phosphate transporters (12).
Several reports have indicated that glutamate is transported in its negatively charged form, in which its carboxyl side chain is deprotonated (13,14). The main evidence comes from mutagenesis studies, implying a conserved arginine residue in the binding site, Arg-446 (EAAC1 sequence numbering) in the formation of a salt bridge with the ␥-carboxylate of the glutamate substrate bound to the transporter (15). This idea was confirmed by the structure of the archeal glutamate transporter homologue Glt Ph (16). In this structure, the bound substrate, aspartate, is located in a binding site that is protected from the extracellular side of the membrane by hairpin loop 2 (HP2), 3 thus locking the substrate into the binding site. Access to the binding site appears to only be possible when HP2 opens, allowing the substrate to bind and/or dissociate (16). This structural picture suggests that substrate binding and/or occlusion occurs partway through the membrane electric field, which would result in electrogenic binding of a negatively charged amino acid. However, to date, no experimental evidence for electrogenic substrate binding has been obtained.
In this report, we provide results from a mutant glutamate transporter, in which substrate translocation and/or Na ϩ binding to the glutamate-bound form is slowed down but still allows for glutamate binding. The mutated amino acid residue, A334E, is located at the tip of HP1 (Fig. 1A), which was previously implied to act as the internal gate (16) and is also important in Na ϩ coordination. The mutation was originally designed to test structural changes involving HP1 during translocation, by engineering charge movement through the introduction of negative charge. However, our results suggest that the charge-altering mutation electrostatically interferes with structural changes and sodium binding to the glutamate-bound transporter, allowing us to isolate substrate binding, and steps associated with it, along the time axis. We demonstrate outward charge movement upon rapid glutamate application to the A334E mutant transporter, which is consistent with an electrogenic glutamate binding step and/or structural changes induced by glutamate binding. The experimental results are compared with computational analysis of the electrostatics of glutamate binding/HP2 closure, which also suggest that binding is voltage-dependent. Kinetic and structural models are proposed to explain the data and to exclude other possible mechanisms of glutamate-induced outward currents.

Experimental Procedures
Molecular Biology, Cell Culture, and Transfection-HEK293 cells (American Type Culture Collection number 1573) were cultured as described previously (6). Subconfluent cell cultures were transiently transfected with EAAC1 cDNA inserted into a pBK-CMV-expression plasmid by use of JETPRIME transfection reagent, following the protocol provided by the manufacturer. The cells were then used for electrophysiological measurements 24 h after transfection. Site-directed mutagenesis was performed using the QuikChange method, as detailed in the manufacturer's protocol (Stratagene).
Whole-cell Current Recording-Whole-cell currents from EAAC1-expressing cells were recorded using an Adams & List EPC7 amplifier under voltage clamp conditions (17). The typical resistance of the recording electrode was 2-3 megaohms; the series resistance was 5-8 megaohms. For recordings performed under steady-state conditions, the glutamate-induced currents were small (typically Ͻ500 pA). Therefore, series resistance was not compensated. For voltage jump experiments, series resistance compensation of 40 -50% was applied in order to accelerate the capacitive charging of the membrane in response to the changes of the membrane potential (18).
All currents were low pass-filtered at 1-10 kHz (KROHN-Hite 3200) and digitized with a digitizer board (Axon, Digidata 1200) at a sampling rate of 10 -50 kHz, which was controlled by software (Axon PClamp). The experiments were carried out at room temperature and physiological pH.
The rapid solution exchange experiments were carried out as described previously (19). Substrates were applied to the EAAC1(A334E)-expressing cell through a quartz tube with an opening of about 350 m that was positioned ϳ5 mm from the cell. The linear flow rate of the solutions emerging from the tube was about 5-10 cm/s, resulting in typical rise times of the whole-cell current of 30 -50 ms.
Ionic Conditions for Rapid Solution Exchange-To test anion-conducting properties, SCN Ϫ was used because it enhances glutamate transporter-associated currents (5). The presence of permeating anions like SCN Ϫ has previously been shown to permanently activate an anion current under homoexchange conditions (6). The anion current was useful in studying this transporter because the A334E mutation interferes with steady-state glutamate transport. To investigate the glutamate concentration-response relationship of the mutant transporter at steady-state conditions, the pipette solution contained 500 M glutamate, 120 mM NaSCN, 2 mM MgCl 2, 5 mM EGTA, and 10 mM HEPES (adjusted with NaOH to pH 7.4). The extracellular solution contained varying concentrations of up to 500 M glutamate, 140 mM NaMES, 2 mM CaMES, 2 mM MgMES, 10 mM HEPES (adjusted with NaOH to pH 7.4). These same conditions were used to test the effect of the mutation on the anion current of the transporter using voltage jump protocols as specified under "Voltage Jumps." Nonspecific currents in the presence of a 200 M concentration of the non-transportable competitive inhibitor DL-threo-␤-benzyloxyaspartic acid (TBOA) (20) were subtracted to obtain the A334E-specific component.
To study the effect of the A334E mutation on the coupled transport current of the transporter, the pipette solution contained 140 mM KMES, and the extracellular solution contained 500 M glutamate, 140 mM NaMES, 2 mM CaMES 2 , 2 mM MgMES 2 , 10 mM HEPES (adjusted with NaOH to pH 7.4) with and without 200 M TBOA. To select cells with transportermediated anion conductance, 140 mM NaSCN was used in place of NaMES because the transporter activity of the mutant is almost negligible. The transporter activity was also investigated under the Na ϩ /glutamate homoexchange conditions; the pipette solution contained 500 M glutamate, 140 mM NaMES, 2 mM CaMES 2 , 2 mM MgMES 2 , and 10 mM HEPES (adjusted with NaOH to pH 7.4), whereas the extracellular solution contained 500 M glutamate, 2 mM MgMES 2 , 5 mM EGTA, and 10 mM HEPES (adjusted with NaOH to pH 7.4) with and without 200 M TBOA. An extracellular solution containing 140 mM NaMES in place of 140 mM NaSCN was used to test for transporter activity. The Na ϩ /glutamate homoexchange mode makes use of the saturation of the Na ϩ and glutamate binding sites both intra-and extracellularly, making the potassium binding sites inaccessible; therefore, it is not associated with steady-state transport current.
Voltage Jumps-Voltage command pulses in 20-mV increments were 20 ms in duration from Ϫ100 to ϩ80 mV from a holding potential of 0 mV. Because most cell systems contain intrinsic, voltage-sensitive channels, transporters, or pumps that lead to nonspecific transient currents, TBOA, a non-transportable transporter inhibitor, was used to detect specific transient currents induced by change in membrane potential (6,21).
Laser Photolysis-Laser pulse photolysis experiments were performed as described previously (6). 4-Methoxy-7-nitroindolinyl (MNI)-caged glutamate (TOCRIS) was applied to the cells using a small quartz tube and photolyzed with a light flash (355 nm Nd:YAG laser, Minilite series, Continuum, Santa Clara, CA) delivered through an optical fiber (350-m diameter). The MNI-caged glutamate solutions were freshly prepared before starting each experiment. The yield from the photolysis of 1 mM MNI-caged glutamate was calibrated to release 200 M free glutamate for the maximum laser energy, according to the procedure described previously (22). Neutral density filters were applied to adjust the laser energy and attenuate glutamate release. Before and after laser-pulse photolysis experiments, a standard glutamate concentration of 100 M was applied to the cell by solution exchange to determine the concentration of glutamate released with the laser flash. After photolytic glutamate release, the steady-state current amplitude was compared with the steady-state current amplitude from the calibration experiment (100 M glutamate) and the known concentrationresponse curve, yielding the released concentration. MNIcaged alanine was kindly provided by Dr. John Corrie and prepared as described previously (11).  (23) and subsequent transfection with pBluescript KS(Ϫ) harboring EAAC-1-WT or -A334E or with the vector alone, as well as L-[ 3 H] glutamate transport, was done as published previously (24).
Data and Statistical Analysis-The non-linear regression fits of the experimental data were performed using Origin (Microcal Software, Northampton, MA) or Clampfit (pClamp8 software, Axon Instruments, Foster City, CA). Concentration-response data were described with a Michaelis-Menten type relationship.
Experiments were performed in triplicate from at least two different cells. Error bars represent S.D.
Electrostatic Calculations-We have used the adaptive Poisson-Boltzmann solver, APBS (25), together with the APBSmem Java routines (26) to calculate the electrostatic energies of the glutamate transporter embedded into an implicit membrane in various states (HP2 open and closed, substrate bound and unbound), as described (27). In brief, the following modified version of the linearized Poisson-Boltzmann equation was used.
Ϫٌ͑⑀͑r ជٌ͒͑r ជ͒͒ ϩ 2 ͑r ជ͒͑r ជ͒ ϭ e4 Here, ⑀ is the dielectric constant, which depends on the spatial coordinate, ϭ e⌽/k b T, where ⌽ is the electrostatic potential, e is the elementary charge, T is the temperature, and k b is the Boltzmann constant. is the Debye-Hückel screening constant, and is the charge density. f(r) is the Heaviside step function, set to 1 in the intracellular solution and 0 in the membrane, protein, and extracellular solution.
The total electrostatic energy, E, is then computed by summing up over the product of the local charge and the potential (28), where dV is the volume element. Plotting ⌬E between different states versus the membrane potential yields the valence from the slope (29).
Molecular Dynamics (MD) Simulations-The model system for MD simulations was set up using VMD (30), by placing an EAAT3 homology model based on the Glt Ph structure (Protein Data Bank code 2NWX (8); 2 Na ϩ ions and 1 aspartate bound to each subunit) into a pre-equilibrated POPC (1-palmitoyl-2oleoylphosphatidylcholine) lipid bilayer with the dimensions 130 ϫ 130 Å. Details can be found in Ref. 18. The total system contained 146,859 atoms. The system was equilibrated for 4 ns at a constant pressure of 1 atm and a constant temperature of 310 K, using Langevin dynamics. For the following production runs, we used NVT conditions (constant number of particles, volume, and temperature (canonical ensemble)). The cut-off for short range interactions was 12 Å. For electrostatic interac-tions, we used the particle mesh Ewald method (31) implemented in NAMD. Bonds, including hydrogen atoms and TIP3P water, were kept rigid using SHAKE. The time step of the simulations was 2 fs (NAMD) or 4 fs (ACEMD (32)). All simulations were performed using the CHARMM27 force field (33).

Rapid Glutamate Application Induces Outward Transient
Currents in EAAC1(A334E)-The functionality of the mutant glutamate transporter was assessed by following relaxation of currents in response to rapid application of the transported substrate, glutamate. Glutamate was applied by photolysis of caged glutamate (release time constant of 200 ns (34)), providing a time resolution in the 100 s time range. After photolytic release, glutamate binds to the transporter, and the subsequent transporter reaction steps are synchronized in time, providing time-resolved information on the charge movement associated with these steps.
As shown in Fig In contrast, rapid glutamate application induced transient inward current in EAAC1(WT) (Fig. 1C) due to the inward movement of positive charge. It was previously suggested that this inward transient current in the wild-type transporter, which displays rising and decaying phases, is caused by the following molecular events following glutamate release (6, 35-36): 1) rapid rise of the current due to diffusion-controlled glutamate binding and 2) biexponential decay of the current caused by the binding of Na ϩ to the empty transporter, electrogenic glutamate translocation across the membrane, and electrogenic Na ϩ binding to the glutamate-bound form of the transporter and/or a conformational change associated with this binding. All of these events cause the movement of positive charge deeper into the transmembrane electric field until a new steady state is reached, thus generating inward transient current. A typical trace is shown in Fig. 1C, and average currents are shown in Fig. 1D.
The glutamate concentration dependence of the relaxation rate constant, 1/, of the decaying phase of the current is shown in Fig. 1F, allowing insights into the molecular basis underlying the transient outward current. At low concentrations, the 1/ versus [Glu] relationship is linear. Such a linear concentration dependence is expected for a diffusion-controlled substrate binding reaction with a bimolecular rate coefficient of 2 ϫ 10 7 M Ϫ1 s Ϫ1 (Fig. 1F, dashed line), as proposed previously for EAAC1(WT) and other subtypes (35,37). However, at high [Glu], deviation from linearity is observed, with 1/ leveling off (Fig. 1F, solid line), indicating that glutamate binding is not rate-limiting under these conditions. Leveling off is expected if a reaction other than binding (e.g. a conformational change) follows the rapid substrate binding process.
The voltage dependence of the relaxation rate constant for the rising and decaying phases of the outward transient current is shown in Fig. 1G. The rate constant associated with the decay is accelerated slightly at positive potentials, as predicted for a process in which negative charge moves into the membrane dielectric (e.g. glutamate binding). The weak voltage dependence of this rate constant may be caused by other, electroneutral steps partially limiting the rate of the outward current decay.
Several interpretations (Fig. 1E) are possible to explain the glutamate-induced outward current, including 1) inward movement of the negative charge of Glu-334 (it should be noted that the pK a value of the Glu-334 side chain is not known, so it is also possible that the side chain is protonated and, thus, neutral); 2) inward movement of negatively charged substrate, glutamate; and 3) inward movement of negatively charged anions from the extracellular solution. The experiments and computational analysis described below will differentiate between these possibilities.

Computational Analysis Suggests That Glutamate Binding Is
Electrogenic-To test whether electrogenic glutamate binding (mechanism 2 in Fig. 1E) could be responsible for the observed outward charge movement in EAAC1(A334E), we performed electrostatic calculations (Fig. 2, A and B) based on structural models of the open loop, glutamate-free and closed loop, glutamate-bound configurations ( Fig. 2C; based on EAAC1 homology models generated using Glt Ph as a template (18,27)). Solving the Poisson-Boltzmann equation allows the computation of the electrostatic energy of the two states (open loop and closed loop/glutamate-bound), which yields the valence of the transition, when computed as a function of the membrane potential ( Fig. 2A). The results (Fig. 2B) suggest that substrate binding, combined with loop closure, is associated with a negative valence of Ϫ0.27 for both EAAC1(WT) and EAAC1(A334E). In contrast, loop closure alone is predicted to be associated with little charge movement (Fig. 2B) for the wild-type transporter and a small negative valence for EAAC1(A334E). These calculations suggest that electrogenic binding of negatively charged glutamate, possibly combined with HP2 loop closure, may be the underlying mechanism for the observed outward current in  Charge Movements in Response to Voltage Jumps-If substrate binding is electrogenic, it should be possible to perturb the binding equilibrium with step changes of the membrane potential. Consistent with this idea, transient currents were observed in EAAC1(A334E)-transfected cells in response to voltage jumps (Fig. 3). TBOA was used to isolate the EAAC1specific component of the currents by subtracting currents arising from other nonspecific transport activity and those arising from charging and discharging the membrane capacitance (6, 21) (control; Fig. 3B). In the presence of glutamate, these currents were mainly transient in nature (Fig. 3A; the small steadystate component was caused by instabilities in the cells toward the solution exchange). Integrating the off-current response yielded the charge. The charge as a function of voltage was fitted to a Boltzmann-like function (Fig. 3C). The resulting apparent valence of the A334E mutant transporter (0.67) was comparable with that of the wild-type transporter (0.64). These results suggest either that electrogenic equilibration of the glutamate translocation step still takes place (see below for more experiments that negate this possibility) or that electrogenic glutamate binding causes the voltage-induced charge movement. It should be noted that the voltage jump analysis does not provide information on the direction or the sign of the charge movement.
It has been previously shown that binding of Na ϩ to the glutamate-free transporter or a conformational change linked to it can result in charge movement in the wild-type transporter (6,21). This charge movement is blocked by the non-transportable inhibitor TBOA (38). In agreement with this hypothesis, the glutamate-free transporter showed specific TBOA-sensitive transient currents upon stepping the membrane potential in the presence of Na ϩ (Fig. 3D). In this case, the apparent valence was 0.42 ( Fig. 3, E (control) and F), similar to that of the wildtype transporter. This result indicates that binding of Na ϩ to the empty transporter is not altered by the A334E mutation. In fact, Na ϩ binding is a prerequisite for glutamate binding (6). Therefore, based on the results showing glutamate interaction with the mutant transporter, it was not expected that Na ϩ binding to the empty transporter would be affected.

Na ϩ -dependent Leak Anion Currents and Glutamate-induced Anion Currents Are Still Functional in the A334E Mutant Transporter, but the Activation of the Anion Current Is Slowed
Significantly-To test whether the A334E mutation interferes with the anion conductance of the glutamate transporter, we carried out recordings in the anion conductance mode. These experiments were performed because it is known that some mutant transporters lack transport activity but can still support anion current (39,40). We used SCN Ϫ as the permeating anion because the glutamate transporter shows a high conductance for hydrophobic anions, such as thiocyanate (5). This permeability is further enhanced by the transported substrate (5). The wild-type transporter facilitates leak anion currents in the presence of SCN Ϫ and Na ϩ ; this current can be inhibited by nontransportable inhibitors, such as TBOA (35,41). Fig. 4A shows that Na ϩ -induced leak anion currents are also present in EAAC1(A334E). These currents are directed inward, due to outflow of internal SCN Ϫ through the Na ϩ -induced leak anion conductance, which can be inhibited by TBOA.
In the presence of substrate, Na ϩ , and a permeable anion, glutamate transporters facilitate what is referred to as a glutamate-gated anion conductance. Glutamate has to bind to the transporter for its activation, but the conductance is not thermodynamically coupled to glutamate transport. This anion current strictly follows the driving force for anions (5,35). Results shown in Fig. 4B illustrate that EAAC1(A334E) exhibits a glutamate-gated anion conductance in the homoexchange mode (glutamate was present on both sides of the cell membrane). In the presence of intracellular thiocyanate, the anion current was inwardly directed, and its magnitude increased in the presence of glutamate, as supported by previous studies on the wild-type transporter (5). The voltage dependence of this anion current (shown in Fig. 4, C and D) is comparable with that observed previously in the wild-type transporter (42). Subtraction of current in the presence of TBOA was used to isolate the specific current (Fig. 4, A and B). EAAC1(A334E) also exhibited anion currents with SCN Ϫ in the extracellular rather than in the intracellular medium. Now an outward current that was sensitive to TBOA was generated (not shown), consistent with uncoupled anion flow into the cell. Together, these results indicate that the A334E mutation does not substantially inhibit the anion conductance of the transporter.
The glutamate transporter anion conductance can be activated by application of glutamate. In the wild-type transporter, the anion conductance is activated by glutamate with a time constant of 0.9 Ϯ 0.15 ms (Fig. 4F, gray trace). In contrast, the activation of the anion conductance by glutamate is dramatically slowed in EAAC1(A334E), with a time constant, , of 1000 Ϯ 200 ms (Fig. 4E, black trace; WT control is shown for comparison in gray), indicating a 1000-fold slower activation of the anion conductance than in the wild-type transporter. Fig.  4F shows the activation of the EAAC1(A334E) anion current (black trace) with high time resolution (millisecond time scale), demonstrating little activation within the first 60 ms after glutamate release, in contrast to the wild-type transporter (gray). The time course of the activation of the anion conductance by glutamate is thought to reflect the glutamate-induced increase of the population of the intermediate occluded state (43,44) along the translocation pathway, suggesting that the rate of formation of this state is severely inhibited by the A334E mutation.
The Apparent Affinity for Glutamate Is Not Significantly Affected by the A334E Mutation- Fig. 5A shows a typical current recording from the A334E mutant transporter when glutamate was applied by rapid solution exchange in the Na ϩ / glutamate homoexchange mode. The intracellular solution contained SCN Ϫ , and the inwardly directed currents arise from the anion's exit from the cell (V ϭ 0 mV). The amplitude of the anion current was dependent on the glutamate concentration in a Michaelis-Menten-like fashion (Fig. 5B). The apparent affinity (K m ) for glutamate was determined to be 30 Ϯ 5 M, whereas that of the wild-type transporter was 19 Ϯ 8 M. These results show that the mutation did not dramatically alter the apparent affinity of the transporter for the substrate. Therefore, the dramatic changes in functional behavior of the mutant transporter are not caused by changes in the substrate binding pocket.
Coupled Steady-state Transport Current Is Not Observed in the A334E Mutant Transporter-Wild-type glutamate transporters catalyze significant voltage-dependent inward transport current, caused by the inward movement of two positive charges (4) for each transported glutamate anion (Fig. 6B). In contrast, EAAC1(A334E)-expressing cells displayed little, if any, transport current when glutamate was applied to the extracellular side in the presence of internal K ϩ (Fig. 6A, forward transport conditions; glutamate was applied as indicated by the gray bar and removed at t ϭ 4 s) at 0 mV as well as at increased driving force (up to Ϫ100 mV; Fig. 6B). In addition, [ 3 H]glutamate uptake is strongly inhibited by the mutation (Fig. 6C). Together, these results suggest that glutamate translocation and/or K ϩ -induced relocation is slowed.
Poisson-Boltzmann analysis of the membrane potential distribution along the membrane normal for an EAAC1 homology model, based on Glt Ph , suggests that the 334 side chain moves through a major proportion of the transmembrane electric field during the outward to inward facing transition (Fig. 6, D and E, position of A334E indicated by arrows, membrane potential color-coded from 0 mV (blue) to Ϫ100 mV (red)). Whereas the valence associated with this transition is positive for the wildtype transporter (18), as determined from experiments as well as computation (z ϭ ϩ0.15; Fig. 6F), the computed valence for the mutant transporter is large and negative (z ϭ Ϫ0.71 (Fig.  6F) if the Glu-334 side chain is deprotonated), implying that the large alteration of the electrostatic balance of the translocation process is likely to inhibit both the translocation and the K ϩ -induced relocation reactions.

Molecular Dynamics Simulations of Wild-type and Mutant
Transporter-To test the effects of the A334E mutation in silico, we performed 1-s MD simulations (Fig. 7). The structures at the end point of the simulations for wild-type and mutant EAAC1 are shown in Fig. 7 (A and B). The mutant transporter showed significant distortion of the structure at the tip of HP1 (Fig. 7, A and B), whereas the substrate binding site and HP2 structures were virtually unchanged (not shown). Consistent with these findings, the substrate was stable in the EAAC1(A334E) and wild-type binding sites for 100 ns up to 1 s (Fig. 7, C and D), consistent with the experimental data showing little effect of the mutant on substrate apparent affinity. To quantify the effect of the A334E mutation on the structure at the tip of HP1, we determined the radial distribution function of the distance of water from the C␤ atom of Ala-334 and A334E (Fig. 7, E and F). The results show that water penetration to the tip of HP1 is significantly increased in the mutant transporter, possibly caused by the introduction of a potential negative charge in this position. This structural distortion could explain the inability of the transporter to catalyze translocation while preserving the ability to bind the substrate.
Sodium ions in site Na1 and the predicted Na3Ј site were stable in all simulations performed and did not dissociate within 1 s (data not shown). These results are consistent with the voltage jump analysis, indicating that the A334E mutation does not interfere with Na ϩ binding to the empty transporter. However, sodium ions were unstable in simulations of the Na2 site of the mutant transporter, dissociating within nanoseconds of the start of the simulation. Because binding of sodium to the Na2 site is thought to be required for the closure of the HP2 loop (8), it is possible that defective Na ϩ binding to the loaded transporter is the reason for the inability of the transporter to catalyze glutamate translocation.

Sodium Dependence of the Transient Outward Current-
Upon close inspection, the transient outward current observed after rapid glutamate application to EAAC1(A334E) is followed by a minor but significant inward component (Fig. 8A). This result raises the possibility that some electrogenic reactions, such as Na ϩ binding to the empty transporter, may still be func-  tional after mutation, as indicated by our results shown above. To test this hypothesis, we repeated the [glutamate] jump experiment at lower extracellular [Na ϩ ] (20 mM; Fig. 8A), shifting the Na ϩ binding equilibrium further to the non-bound state. Interestingly, most (but not all; Fig. 8B) of the outward charge movement was eliminated, but a sizable inward current was observed. This current decayed significantly slower than the outward current at 140 mM Na ϩ (Fig. 8C). These results suggest that electrogenic Na ϩ rebinding to the empty transporter, which is still functional in EAAC1(A334E) (Fig. 3), starts masking the glutamate-induced outward charge movement at low sodium concentrations.
Binding of a Neutral (Zwitterionic) Substrate Does Not Induce Outward Charge Movement-A glutamate transporter with the mutation R446Q was shown previously to bind neutral amino acids, such as glutamine and alanine, instead of the negatively charge glutamate. We exploited this mutation to test whether neutralizing the substrate would eliminate the outward charge movement. The results of rapid alanine application to the EAAC1(R446Q,A334E) double mutant transporter are shown in Fig. 9. Rapid alanine application (300 M) by photolysis of caged alanine resulted in a transient inward current (Fig. 9A), in contrast to the outward current observed upon glutamate application to EAAC1(A334E). This alanine concentration is close to saturating, as demonstrated by the concentration-response curve shown in Fig. 9B. These results demonstrate that  a negatively charged substrate is necessary to generate transient outward current. They also suggest that the outward current is not caused by inward movement of the Glu-334 side chain in the mutant transporter, which is preserved in the double mutant transporter.

Discussion
The major finding of this study is that binding of negatively charged substrate to the glutamate transporter EAAC1 and/or a conformational change closely associated with binding is electrogenic and, thus, is associated with apparent outward charge movement, caused by inward movement of a negative charge (Figs. 1B, 2, and 6A). This charge movement was uncovered in experiments with a transporter with an Ala to Glu mutation in position 334, located at the tip of hairpin loop 1. The results suggest that the outward charge movement is hidden in the wild-type transporter because it is overcompensated by simultaneous inward movement of positive charge from glutamate translocation and Na ϩ binding to the empty and substratebound transporter. This is an important finding because glutamate binding adds to the substantial list of electrogenic partial reactions in glutamate transporters (6,27,45) in which charge movement within the membrane is distributed over several reaction steps. This includes Na ϩ binding, dissociation, K ϩ -induced relocation, and substrate translocation steps (27,46).
Based on the current knowledge of the alternating access, elevator-like transport model of glutamate transporters (47,48), the existence of an electrogenic substrate binding process is not unexpected. In current models, substrate binds to the open loop configuration, in which HP2 is open to the extracellular side (49,50). Upon binding, closure of HP2 can occur, generating a complex that can undergo the structural changes associated with the translocation reaction. Glutamate, in this occluded complex, is buried more deeply within the membrane, thus generating charge movement. Our electrostatic calculations (Fig. 2) suggest that the binding process occurs within part of the transmembrane electric field, whereas the HP2 closure process is associated with little, if any, voltage dependence.
We performed kinetic simulations to provide a quantitative explanation of the experimental results. The simulations are based on a sequential Na ϩ substrate-Na ϩ binding model (Fig.  10), which has been proposed in previous work (6,51,52). We have also included a reaction step following substrate binding, possibly a conformational change, evidence for which has been obtained (6,40). The insets shown in Fig. 10 illustrate that such a model can reproduce the experimental data ( Fig. 1) well, when assuming that Na ϩ binding to the glutamate-bound transporter and translocation are inhibited by the A334E mutation, whereas the kinetics of Na ϩ binding to the empty transporter and glutamate binding remain unchanged. Inward movement of negative charge (valence ϭ Ϫ0.16) was modeled into the glutamate binding step. The detailed kinetic parameters, which were not fitted but rather estimated based on the present and previous work are listed in the legend of Fig. 10. The possibility cannot be excluded that other sets of parameters can also account for the experimental data, including negative valence modeled for the step following the actual glutamate binding.
The voltage dependence of the apparent affinity of the wildtype transporter for glutamate has been investigated previously at steady state (52). At 100 mM [Na ϩ ], this affinity was found to be only weakly dependent on the membrane potential, slightly increasing at positive voltages. Although increased affinity at positive potentials is expected for voltage-dependent glutamate binding, many processes other than glutamate binding affect the experimentally determined apparent affinity at steady state. Most likely, the apparent affinity is mainly controlled by the two slowest steps in the cycle, the voltage-dependent glutamate translocation process and the K ϩ -dependent relocation step (see Ref. 35 for equations). Because both of those processes are accelerated in an almost equal fashion by more negative transmembrane potential, the apparent affinity is relatively unchanged at different voltages. These results further demonstrate that it can be difficult to obtain information on the voltage dependence of non-rate-limiting partial reactions at steady state but that time-resolved, pre-steady-state analysis is necessary to detect electrogenicity, as has been shown for other transporters (53).
Our results allow us to clearly distinguish between the proposed electrogenic substrate binding model and two alternative major mechanisms that could explain the outward charge movement observed in the mutant transporter. The charge movement cannot be caused by chloride (mechanism 3 in Fig.  1E) as a charge-carrying species because it is still present in the absence of chloride in both the intracellular and extracellular solutions (Fig. 1). Interestingly, the A334E mutation does not interfere with the ability of the transporter to support the anion conductance (Fig. 4). This result is somewhat surprising because the side chain of Ala-334 is very close to the residues at the tip of HP1, which have been recently proposed to contribute to anion pore formation (54). It would be expected that an additional negative charge close to those positions would electrostatically inhibit anions from entering the pore. However, our MD simulations suggest that water permeation is increased at the tip of HP1 by the mutation. This effect, possibly enhancing anion permeation, may counteract any negative effects of electrostatics.
The activation of the anion conductance is dramatically slowed in the A334E mutant transporter relative to the wild type. It was proposed that the anion-conducting state corresponds to the intermediate, occluded state observed in a crystal structure (43, 44, 54 -56). If this is the case, transition from the outward occluded state to the intermediate state is slowed by the A334E mutation, consistent with our observation of lack of steady-state transport (Fig. 6). It is interesting to speculate on the structural basis of the defective step of the A334E mutation. In the wild-type transporter, the closure of the external gate/ HP2 loop requires the binding of the "last" sodium to the Na2 site (8), and this is followed by the translocation step thought to occur via an "elevator-like" movement (47,48). Even in the absence of glutamate, sodium can activate the anion conductance. This suggests that the intermediate anion-conducting state (54,55), which requires a partial "elevator" movement of the transport domain, can be reached in the presence of sodium alone. Thus, it is possible that if A334E transporters are defective in the binding of the last sodium to the Na2 site, a partial "elevator" movement through the membrane electric field still can take place, which reflects the electrogenic glutamate binding step. Evidently, such a scenario also accounts for the defective transport by the mutant (Fig. 6).
Our results are somewhat reminiscent of what has been observed in a transporter with another charge-altering mutation, D439N (40), which also showed a dramatic slowdown of anion current activation by glutamate. However, in the D439N mutation, a potential negative charge is eliminated, not introduced, as opposed to the A334E mutant transporter. Rapid application of glutamate induces transient outward current in EAAC1(D439N) (not shown), which displays kinetics similar to those observed in the A334E mutant transporter. Therefore, the molecular basis of the effect of the two oppositely chargealtering mutations, namely the slowing of the Na ϩ -dependent processes of the glutamate-bound transporter, is most likely the same.
The second alternative mechanism is based on inward movement of negative charge in position 334 (mechanism 1 in Fig.  1E). However, in the double mutant transporter A334E,R446Q, which binds neutral instead of acidic amino acids, no outward charge movement is observed (Fig. 9). Instead, inward transient current is observed, which is caused by the inward movement of bound Na ϩ , as in the wild-type transporter and glutamate as substrate. If the transient outward current was caused by inward movement of A334E, it should still be observed in the double mutant transporter. These results directly show that the outward charge movement is not caused by inward movement of the A334E side chain.
Our results indicate that glutamate translocation is impaired by the A334E mutation. In light of the elevator type mechanism for transport, in which the tip of HP1 moves through a large part of the hydrophobic interior of the membrane (see electrostatic calculations in Fig. 6), this result is not surprising. The glutamate side chain in position 334 may add negative charge or at least hydrophilic character to the tip of HP1, raising the possibility that the energy barrier for motion of HP1 across the bilayer is increased.
Is glutamate binding from the intracellular side of the membrane also electrogenic? At present, our data do not provide insights into this question. However, considering the symmetrical nature of the transporter structure and the known electrogenic binding of Na ϩ to its intracellular binding site (46), it is likely that intracellular glutamate binding is also electrogenic. This possibility could be tested by rapidly applying glutamate to the intracellular side of the membrane, which we have done in the past in the wild-type transporter (46). Such experiments FIGURE 10. Sequential Na ؉ /substrate binding model explaining the experimental data. The model is based on previous results (40) as well as current data. T, transporter; S, substrate. The simulations for the wild-type transporter (green box) and A334E (red box) were performed as described previously (6,60), using numerical integration of the rate equations pertaining to the sequential binding mechanism. The kinetic parameters were estimated based on previous results. They do not provide a unique representation of the data but can describe the experimental results well. The following parameters were used at V ϭ 0 mV (wild type): Na ϩ binding to the empty transporter: k for ϭ 20 M Ϫ1 ms Ϫ1 , k rev ϭ 1 ms Ϫ1 , z Q ϭ 0.4; glutamate binding: k for ϭ 10,000 M Ϫ1 ms Ϫ1 , k rev ϭ 1 ms Ϫ1 , z Q ϭ Ϫ0.16; Na ϩ binding to the glutamate-bound transporter: k for ϭ 1 M Ϫ1 ms Ϫ1 , k rev ϭ 0.2 ms Ϫ1 , z Q ϭ 0.6; TNa ϩ S 7 TЈNa ϩ S: k for ϭ 5 M Ϫ1 ms Ϫ1 , k rev ϭ 0.1 ms Ϫ1 , z Q ϭ 0.4; [Glu] ϭ 200 M; [Na ϩ ] ϭ 0.14 M. For the A334E mutant transporter, k for (TNa ϩ S 7 TЈNa ϩ S) was set to 0, with all other parameters kept at the same value. The apparent rate constant of glutamate release from the caged precursor was set to 2 ms Ϫ1 , accounting for the non-instantaneous rise of the outward current.
with the A334E mutant transporter will be performed in the future.
The tip of HP1 has been investigated in previous mutagenesis studies and has been found to be important for substrate interaction as well as K ϩ -induced relocation of the transporter (57). In addition, our results suggest that this region may be also involved in controlling reactions associated with the translocation process.
Glutamate transporters are known to regulate the glutamate concentration in the synaptic cleft after presynaptic release. Due to the relatively slow, millisecond kinetics of glutamate translocation across the membrane, it was hypothesized that rapid buffering of glutamate through binding to the transporter binding site contributes to controlling the synaptic glutamate concentration at short times after release (58). Buffering is possible due to the large density of transporter binding sites in and around the synaptic contacts, which was estimated to about 10,000 sites/m 2 (1). Buffering kinetics were incorporated into simulations of the time dependence of synaptic glutamate concentration (2,59). Interestingly, one of these studies showed a time dependence of the K d of the transporter for glutamate, which decreases over time (2). Thus, buffering occurs with a K d of 100 M directly after release, decreasing to a K d of 5 M at steady state. This suggests that the buffered glutamate concentration also decreases at increasing times after release. The finding of voltage-dependent glutamate binding adds another variable, suggesting that the K d for buffering depends not only on time but also on membrane potential. Thus, the optimum buffering range would be shifted to lower concentrations at depolarized membrane potential. This effect could be particularly important for neuronal glutamate transporters, such as EAAC1, due to the voltage changes in neuronal membranes during synaptic transmission events.

Conclusions
In summary, we have demonstrated that binding of the anionic organic substrate to the neuronal glutamate transporter EAAC1 is electrogenic and, thus, associated with transient outward charge movement when glutamate binds from the extracellular side of the membrane. This finding is important because it further supports previous evidence on the distribution of voltage dependent partial reactions over the transport cycle, resulting in a relatively weak voltage dependence of steady-state glutamate transport. The results also suggest that voltage dependence of the K d for glutamate should be included in future simulations of the time dependence of synaptic [glutamate] after presynaptic release. Our findings may provide insight into substrate binding by other secondary active transporters, in which charged organic or inorganic substrates are transported across membranes.
Author Contributions-R. T. performed experimentation, design, analysis and interpretation of data, and writing of the manuscript. Z. T. performed experimentation and data analysis. N. S. performed experimentation and data analysis. B. K. performed data analysis and interpretation and writing of the manuscript. C. G. performed experimentation, design, analysis and interpretation of data, and writing of the manuscript.