Mechanism of transport modulation by an extracellular loop in an archaeal Excitatory Amino Acid Transporter (EAAT) homolog

Secondary transporters in the excitatory amino acid transporters (EAATs) family terminate


INTRODUCTION
Glutamate is the major excitatory neurotransmitter in the mammalian central nervous system, mediating synaptic transmission by activating receptors in post-synaptic neurons (1). Glutamate is removed by reuptake into glia and neurons by members of the EAAT family, which transport 1 glutamate molecule into the cell coupled to 3 Na + ions and 1 H + accompanied by the countertransport of 1 K + ion (2). EAATs are members of the dicarboxylate/amino acid:cation symporter (DAACS) family (TCDB ID 2.A.23) that also includes members that transport dicarboxylates and other amino acids (3)(4)(5). Glt Ph , from the archaeon Pyrococcus horikoshii, is 30% identical to the EAATs (6), and transports aspartate (L-asp); it is the only DAACS member for which there is high resolution structural information (7)(8)(9)(10). Glt Ph is an excellent model for the EAATs with similar functional properties, including the co-transport of substrate with 3 Na + ions (11,12) and the presence of a thermodynamically uncoupled anion conductance (13). However, Glt Ph does not require K + to recycle the glutamate-free transporter from inward facing to outward facing states, instead recycling in a substrate-and ionfree state (11). Glt Ph , like other members of the EAAT family, is a trimer; each protomer consists of 8 transmembrane helices (TMs) and 2 re-entrant hairpin loops -the tips of which, along with transmembrane helices 7 and 8, form the substrate binding site (7,14) (Fig. 1A). Each protomer is organized into 2 distinct domains; a stationary trimerization domain that mediates all intersubunit contacts, and a mobile transport domain that fully encompasses the substrate binding site (9,15). Crosslinking, structural, computational studies, and more recently, electron paramagnetic resonance (EPR) spectroscopy suggest a model in which the transport domain moves ~18 Å and rotates ~30 o through the trimerization domain (9,(16)(17)(18). In a simple kinetic scheme for Glt Ph , L-asp and 3 Na + ions bind to the outward facing state (OFS) (Fig.  1B, step 1), facilitating this dramatic conformational change and resulting in a pistonlike movement of the substrate binding site across the membrane (Fig. 1B, step 2); in our scheme this step includes the formation of an occluded outward facing state ("closing", perhaps by HP2 closure), the large scale translocation of the domain, and the transition from "inward occluded" to "inward open" substrate-bound states ("opening"). The substrates are released into the cytoplasm from this inward facing state (IFS) (Fig. 1B, step 3) and the empty transport domain transitions from the IFS back to OFS to restart the cycle (Fig. 1B, step 4; note that in the mammalian isoforms K + is bound for this step). This return step also presumably includes "closing" and "opening" steps to permit the translocation step of the occluded apo transporter (whose structure was recently published (19)). The trimerization domain is attached to the transport domain by 3 loops connecting TMs 2-3, 3-4 (3L4) and 5-6 (7). Comparison of the outward-facing and inward-facing crystal structures reveals distinct conformational changes in 2-3 and 5-6 loops; however, 3L4 was either unresolved or involved in lattice-packing interactions, meaning that the true structure of 3L4 is unknown (7,9). The 21 amino acid, proline-rich 3L4 undergoes substrate-dependent conformational changes, as demonstrated by limited trypsin proteolysis analysis and fluoroscein-5-maleimide accessibility experiments (20). Furthermore, cleaving the protein backbone within 3L4, by means of an engineered Factor X protease recognition site, resulted in almost complete loss of transport activity despite the protein maintaining its structural integrity (20). This novel observation revealed that 3L4 plays a critical role in the transport cycle. In this work, we probe the role of 3L4 in the mechanism of L-asp transport by Glt Ph using crosslinking, binding, and transport assays. We demonstrate that cleaving 3L4 specifically inhibits one of the components of the translocation step of the substrate-loaded transport domain of Glt Ph while sparing the equivalent component of the translocation of the substrate-free apo translocation domain, a conclusion with important implications for the overall mechanism of transport.

Experimental procedures Protein expression and purification
All mutations were made using the Quikchange II site-directed mutagenesis kit (Agilent Technologies) and all cysteine substitutions were introduced into a cys-less Glt Ph mutant wherein the single native cysteine had been mutated to serine (C321S), which is fully active (20). Purification of wild-type Glt Ph and its variants was performed essentially as described previously (11). A culture of E. coli TOP10 cells harboring the appropriate expression vector was grown to mid-log phase and expression was induced by addition of 0.1 % (w/v) L-arabinose. Cells were lysed by sonication and membranes were isolated by multiple centrifugation steps. Membranes containing overexpressed, histagged protein were solubilized by addition of ndodecyl Β-D-maltopyranoside (DDM, Anatrace) and incubated for 1 hour at 4 o C with Ni-NTA Superflow resin (Qiagen) pre-equilibrated with buffer containing 20 mM Tris HCl, pH 7.4, 200 mM NaCl, 5 mM L-glutamate, 0.5 mM TCEP and 2 mM DDM. Contaminants were removed by washing with 20 column volumes (CV) of the same buffer plus 40 mM imidazole. Bound protein was eluted by addition of the same buffer plus 250 mM imidazole. The his-tag was removed by overnight, room temperature incubation with thrombin in a ratio of 10 U/mg protein. Thrombin digestion was quenched by addition of 10 mM EDTA and 1 mM AEBSF.

Protein reconstitution
Reconstitution of Glt Ph was achieved using the process pioneered by Rigaud and coworkers (21). Pre-formed liposomes were partially solubilized with Triton X100 (TX100), protein was added resulting in comicellization of the constituents, and detergent was removed using polystyrene beads (Biobeads SM-2). After obtaining inconsistent transport rates, we thoroughly examined the variables involved and found that to obtain high activity proteoliposomes we needed to a) add TX100 from a 10% (w/v) solution of instead of adding from the 100% detergent stock and b) use the light scattering curve to determine the correct detergent ratio for each batch reconstitution rather than using a fixed, predetermined TX100:lipid ratio. A 3:1 ratio of E. coli polar lipid extract and 1-palitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC, Avanti Polar Lipids) are mixed, dried, and resuspended to 10 mg/ml in Inside Buffer (20 mM Tris/HEPES, pH 7.4, 1 mM NaCl and 199 mM KCl), unless indicated otherwise. The lipid suspension was freezethawed 5 times, extruded and then diluted to 4 mg/ml. Aliquots of 10% TX100 were added to the liposomes and incorporation was monitored using absorbance at 540 nm. Upon saturation of the liposomes with TX100, purified protein was added in a ratio of 3 µg protein/mg lipid. Detergent was removed by multiple additions of Biobeads SM-2 polystyrene beads (Bio-rad). Liposomes were separated from Biobeads and collected by ultracentrifugation. Proteoliposomes were resuspended in Inside Buffer to a final concentration of 10 mg/ml (lipid content), snap frozen and stored at -80 o C.

Transport assays
Proteoliposomes were thawed, extruded through a 400 nm filter, collected by ultracentrifugation and resuspended to a final lipid concentration of 100 mg/ml. In a standard transport assay, the reaction was initiated by diluting proteoliposomes 150-fold into reaction buffer containing 20 mM Tris/HEPES, pH 7.4, 100 mM KCl, 100 mM NaCl, 1 µM valinomycin and 100 nM [ 3 H]-L-asp at 30 o C. 200 µl samples were taken periodically and quenched by addition of 2 ml ice-cold quench buffer containing 20 mM Tris/HEPES, pH 7.4, 200 mM LiCl. The quenched reaction was applied to a nitrocellulose filter (Millipore) over a vacuum manifold and then washed with 2 ml of quench buffer. The nitrocellulose filters were dissolved with 3 ml FilterCount liquid scintillation cocktail (PerkinElmer) and radioactivity was counted using Trilux beta counter (PerkinElmer). The solute counterflow assay was performed in the same way except the proteoliposomes were loaded with buffer containing 20 mM Tris/HEPES, pH 7.4, 100 mM NaCl and 1 mM L-asp by multiple freeze/thaw cycles and extrusion through 400 nm filter. The reaction buffer contained 20 mM Tris/HEPES, pH 7.4, 100 mM NaCl and 100 nM [ 3 H]-L-asp.

Factor X digestion
Purified Glt Ph variants were exchanged into Factor Xa digestion buffer (10 mM Tris/HEPES, pH 8, 5 mM CaCl 2 , 100 mM NaCl and 2 mM DDM). Factor X (New England Biolabs) was added in a ratio of 125 µg/1 mg protein and incubated at 37 o C for 24 hours. To discourage the stabilization of the IFS by disulfide formation between 55C and 364C during the 24 hour incubation, where applicable, we added 1 mM L-asp to the reaction. Factor X digestion was stopped by addition of 10 mM EDTA and 1 mM AEBSF. Control samples that were not digested with Factor X were treated in an identical fashion.

Filter binding assays
Following purification by immobilized metal affinity chromatography and digestion by Factor X (both described previously), Glt Ph Xa was further purified and exchanged into binding buffer (20 mM Tris/HEPES, pH 7.4, 200 mM choline chloride, 10 mM NaCl, 1 mM DDM) using size exclusion chromatography (Superdex 200 10/300 GL). 0.2 µM protein was incubated with a range of [ 3 H]-L-asp concentrations (made up in the same buffer) for 2 hours at room temperature. Binding was quenched and protein was precipitated by addition of ice-cold 50% ammonium sulfate. Precipitated protein was applied to a nitrocellulose filter over a vacuum manifold and washed with 2 ml 50% ammonium sulfate. The filters were dissolved in 3 ml FilterCount liquid scintillation cocktail (PerkinElmer) and radioactivity was counted using Trilix beta counter (PerkinElmer). Background binding of radiolabelled substrate was measured by performing the same experiment in the absence of protein. Binding curves are fit to a single site saturation model; were performed at room temperature. Crosslinking timecourse data was fit to a single exponential model described by the following; where, Y 0 is the Y value at time zero, is the amplitude of the curve, is the rate constant. We first set out to explore the effects of 3L4 cleavage on protein conformational sampling by using Hg 2+ -based crosslinking; a wellestablished method used successfully with Glt Ph (8,9,22). However, not only are false positives an issue with Hg 2+ crosslinking, wherein Hg 2+induced protein band shifts are observed when no crosslink is formed (8), but it is also difficult to quench the Hg 2+ crosslinking reaction with commonly used quenching compounds, such as N-ethyl maleimide (NEM) or S-methyl methanethiosulfonate (MMTS). In the absence of a quenching reagent, when attempting to crosslink cleaved Glt Ph in detergent solution, we observed protein bands on SDS-PAGE gels corresponding to crosslink formation between intra-protein fragments and also, surprisingly, inter-subunit fragments. We reasoned unwanted crosslinking was occurring between unreacted cysteines and residual Hg 2+ ions in the quasidenatured environment of the SDS-PAGE gel or sample buffer. This led to difficulties in differentiating genuine intramolecular crosslink products from intact protein and crosslinking artifacts. We therefore switched to using the homobifunctional thiol-reactive reagent, 1,1-Methanediyl Bismethanethiosulfonate (MTS-1-MTS, linker length ~5 Å), which we could effectively and rapidly quench with an excess of MMTS. Indeed, pre-treatment of detergentsolubilized Glt Ph XaC2 with MMTS prevented any crosslinking from taking place between 55C and 364C even in intact protein.

Densitometry
Densitometric SDS-PAGE band analysis was performed using Alphaview software (Cell Biosciences). Each protein band was selected using identically sized boxes and background values were obtained by selecting an identically sized box from a blank region of the gel directly below the band. Values were normalized to a control protein band of known amount.

Cleaving 3L4 does not prevent Glt Ph from sampling the inward-facing conformation
The observed reduction in transport activity by Glt Ph cleaved at residue 125 in the 3L4 may be due to a decrease in the protein's ability to sample particular states in the transport cycle, to changes in its ability to bind substrate or Na + , to changes in transition rates between two or more cycle states, or a combination thereof. We sought here to distinguish among these possibilities. The transition of the Glt Ph transport domain between the outward-and inward-facing conformations is a major conformational rearrangement required for alternating access in this protein (9). To determine whether cleaving 3L4 restricts Glt Ph 's access to one key state, the IFS, we introduced cysteines at positions known to form crosslinks only in that state (9). If such crosslinks form effectively after cleaving 3L4, Glt Ph must be able to sample the inward-facing conformation despite loop scission. Using a gelbased crosslinking assay, we monitored IFS formation in purified, detergent-solubilized, Glt Ph by introducing cysteines at positions 55 and 364 into a cys-less background with a Factor X site at residue 125 (from now on referred to as Glt Ph XaC2, Fig. 1A). Cysteines in these positions were first shown to form a disulfide bond during an electrophysiological characterization of EAAT1 (23). Subsequently, an x-ray structure of the crosslinked form of the equivalent Glt Ph mutant showed that cysteines in these positions are only in close proximity in the IFS (IFS Cα-Cα distance = 7.3 Å, OFS Cα-Cα distance = 27.6 Å) (9). Furthermore, stabilizing the IFS of Glt Ph by crosslinking residues 55 and 364 increases the protein's electrophoretic mobility, making it easily discernible from noncrosslinked protein on a polyacrylamide gel (9). Here, crosslinking was performed in the absence of substrate, so the inward-facing state stabilized will be predominantly empty transporter (Fig.  1B, black box). Initial attempts to crosslink cleaved Glt Ph XaC2 with Hg 2+ were complicated by extraneous reactions with the metal (see Experimental Procedures), so we developed an alternative approach based on the homobifunctional thiol-reactive reagent, 1,1-Methanediyl Bismethanethiosulfonate (MTS-1-MTS, linker length ~5 Å). In the absence of Factor X treatment, the predominant protein band is full length (FL), non-crosslinked Glt Ph XaC2 (Fig. 1C, Lane 1). A small population of crosslinked protein was also apparent, caused by oxidative disulfide formation during the 24 hour incubation, a fact confirmed by the disappearance of this band after reduction with β-mercaptoethanol (β-ME) (Fig. 1C, lane 2). Treatment of Glt Ph XaC2 with Factor X produced two fragments (NTF, Nterminal fragment and CTF, C-terminal fragment) with no full length protein remaining (Fig. 1C, lane 3). Minimal changes in electrophoretic migration of the two fragments upon reduction with β-ME indicate that the protein band positions are not caused by aberrant crosslinking events (Fig. 1C, lane 4). We observed small, crosslink independent differences in all protein band migration profiles in the presence of β-ME. Treating detergentsolubilized Glt Ph XaC2 with MTS-1-MTS alone led to increased electrophoretic mobility, indicative of IFS crosslink formation, and was fully reversible with β-ME treatment (Fig. 1C, lanes 5 and 6). Cleaving Glt Ph XaC2 produces only NTF and CTF fragments, however, if the fully cleaved protein is then treated with MTS-1-MTS, a third, higher molecular weight protein band was also apparent, which is the product of crosslinking the N-and C-terminal fragments (Fig. 1C, lane 7). We confirmed that this band is formed by crosslinking between the proteolytic fragments because, upon reduction, we observed disappearance of the high molecular weight band and an intensification of the N-and Cterminal fragments (Fig. 1C, lane 8). Mutants containing the Factor X site and the single cysteine substitutions did not form this crosslinked band, indicating that both cysteines in the engineered crosslinking pair are required and that intermolecular crosslinking is not responsible (data not shown). This result demonstrates that, despite lacking an intact 3L4, Glt Ph is still able to sample the inward-facing conformation in detergent solution.
In a related assay, where Glt Ph XaC2 is first locked into the IFS by treatment with MTS-1-MTS and then digested with Factor X, we observe full cleavage of 3L4, indicating that 3L4 is accessible to protease in the inward-facing conformation (data not shown). To complement this work, we also made several attempts to stabilize the OFS using engineered cysteines; K55C-G280C, V51C-T275C, L212C-V274C, V216C-A391C, Q220C-I389C, V58C-P283C and L66C-S300C. For all pairs attempted, the introduced cysteine pairs either did not crosslink using MTS-1-MTS (however, one of these examples, L66C-S300C, does crosslink robustly with Hg 2+ ; its structure has been published recently (22)) or a disulfide would form during purification despite efforts to prevent it. These disulfide bonds could only be reduced under harsh conditions that would render the protein inactive.

Cleaved Glt Ph has Na + -driven L-asp transport activity
Having demonstrated that cleaving 3L4 does not prevent Glt Ph from accessing at least one conformation essential to the transport cycle, we reassessed the transport capabilities of cleaved Glt Ph . Following incubation of purified wildtype Glt Ph (Glt Ph WT) and Glt Ph Xa (Glt Ph containing a Factor X recognition site at residue 125) in the presence and absence of Factor X, we reconstituted the treated proteins into liposomes using an optimized protocol (see Experimental Procedures). This protocol increased the substrate uptake approximately fivefold, and thereby permitted detection of transport activity in the Factor Xa cut GltPhXa, activity we could not detect with our previous methods (20). SDS-PAGE analysis of the proteins prior to reconstitution revealed that treatment of Glt Ph Xa with Factor X resulted in complete digestion, whereas, under identical conditions, Glt Ph WT remained intact (Fig. 2B). We monitored initial rates of 3 H-L-asp transport into proteoliposomes containing cleaved Glt Ph in the presence of an inwardly-directed sodium gradient and, contrary to the negligible activity we previously reported, we observed reduced, but still easily detectable activity (~30% compared to wild-type). There were no apparent effects on the activity of Glt Ph WT after treatment with Factor X ( Fig.  2A). Uncut Glt Ph Xa has lower activity (~70%) than Glt Ph WT, which is likely due to the effects of the amino acid substitutions in 3L4 required to introduce the Factor X recognition site and, as demonstrated by the SDS-PAGE analysis, not from cleavage of the protein during expression and purification (Fig. 2B). The ability to support the transport of L-asp, albeit with decreased efficacy, clearly demonstrates that even with a fully cleaved 3L4, Glt Ph can sample all the conformations necessary for transport. The reduced transport rates must therefore stem from either altered thermodynamic properties of one or more states in the cycle or from altered kinetic properties of one or more transitions in the cycle. We sought to establish which part of the transport cycle is compromised upon 3L4 cleavage by systematically isolating steps and transitions of the cycle and observing the impact of 3L4 cleavage at each stage.

Cleavage of 3L4 minimally affects substrate interaction
Closure of HP2, the outer hairpin, is thought to be an essential step in forming the Glt Ph substrate binding site and allowing translocation of the transport domain. Given that 3L4 directly overlies the HP2 hairpin, we considered the possibility that the loop stabilizes the open state of the hairpin, thereby influencing substrate and Na + binding (7,9). If true, this hypothesis predicts that cleavage of 3L4 would substantially affect substrate binding. We tested this model by deriving the kinetic parameters of transport (V max and K m , see Table 1) using functional protein reconstituted into lipid vesicles. The L-asp dose response reveals a modest decrease in the K m of cleaved protein (69.5 ± 14.1 nM compared to 129.5 ± 16.8 nM for Glt Ph WT, Fig. 3A and Table 1), and a large decrease in the V max (9.2 ± 0.7 nmol/mg/min compared to 42.7 ± 4.5 nmol/mg/min for Glt Ph WT, Fig. 3A and Table 1). We observed the same effects of cleaving 3L4 on the Na +dependence of transport: minimal difference to the K m values and a large decrease in the V max (Fig. 3B). As expected, treatment with Factor X had no effect on the K m or V max values for either L-asp or Na + of Glt Ph WT (Fig. 3A and B and Table 1). Glt Ph Xa, in the absence of Factor X treatment exhibits a lower K m compared to wildtype; this is likely due to the 4 amino acid substitution made to create the Factor X site in 3L4 (Table 1 and (20)).
These data reveal that cleaving 3L4 does not substantially affect the K m of Glt Ph for either Lasp or Na + . If 3L4 cleavage had a major impact on substrate binding, K D , we would expect this to be reflected in the K m . The preservation of K m therefore indicates that the substrate binding/unbinding events during the transport cycle are at most minimally affected by 3L4 scission (Fig. 3B, inset, steps 1 and 3). To corroborate this data, we directly measured the binding of L-asp to detergent solubilized protein and calculated the dissociation constants (K D ) for cleaved and intact Glt Ph Xa. Attempts were made to assess L-asp binding using both isothermal titration calorimetry (ITC), a technique recently used to great effect with Glt Ph (22), and equilibrium dialysis. Both methods proved unsuitable for different, technical reasons; the former, due to the large amount of cleaved protein required (the amount of Factor X protease required being the limiting factor) and the latter, due to the inconsistent data presumably caused by the substrate interacting with the dialysis membrane. Ultimately, data obtained using a filter binding assay revealed robust aspartate binding by both intact and cleaved Glt Ph Xa (Fig. 3C). We observed no substantial difference in the K D between intact and cleaved protein (4.0 ± 0.1 µM and 1.4 ± 0.3 µM, respectively) indicating that substrate binding is essentially preserved after 3L4 loop scission. Using this method, we measure a K D of 1.0 ± 0.3 µM for Glt Ph WT (a value in good agreement with the Glt Ph WT K D obtained using ITC under similar conditions (22)) revealing that introduction of the Factor X site into the 3L4 loop of Glt Ph results in a slight decrease in affinity, a decrease that is reversed upon cleavage of the loop (Fig. 3C and Table 1). Thus, the observed decrease in transport activity is not a consequence of decreased affinity of binding rate, but is due to a lower transport turnover rate, highlighting the possibility that cleaving 3L4 impedes other steps in the transport cycle.

The activation energy of transport is increased when 3L4 is cleaved
The activation energy, E a , of a process reflects the relative height of the energetic barrier that must be hurdled for that process to occur. If the effects of 3L4 cleavage are due to alterations in a significant energy barrier in one of the ratelimiting transport cycle reactions, then we would expect cleavage to cause changes in the E a of transport. These changes, in turn, should be reflected in the temperature-dependence of transport. Over a temperature range of 3-40 o C, we observed a ~16 and ~64-fold increase in the initial rates of 3 H-L-asp transport by intact and cleaved Glt Ph Xa, respectively (Fig. 4A). Analyses of these data with an Arrhenius plot revealed a clear temperature dependence for the transport rate and corresponds to an E a of 56.6 kJ.mol -1 and a temperature coefficient (Q 10 ) of 2.13 for intact Glt Ph Xa (Fig. 4B). The Q 10 value and E a for cleaved Glt Ph Xa are ~1.5-fold higher than intact mutant (3.1 and 83.7 kJ.mol -1 , respectively) indicating a higher energy barrier to at least one step in the transport cycle. The activation energy values are composites of the temperature dependence of various conformational transition rates during the transport cycle, but combined with our evidence that the binding reactions are minimally affected by 3L4 cleavage these data point to other steps in the cycle, namely the closing/opening and translocation reactions of loaded and unloaded transport domains. Since our current experimental methods do not have the resolution to distinguish among these steps, we will henceforth collectively refer to them as the "transfer steps" since together they transfer the accessible substrate binding sites from one face of the membrane to the other. These steps together comprise step 2 (the loaded transfer) for the substrate-loaded transporter and step 4 (the apo transfer) for the empty protein in our simplified kinetic scheme ( Figure 1B).

Cleaving 3L4 results in severely diminished substrate exchange
Thus, the transfer reactions are groups of conformational changes that alternately expose the substrate binding site from one side of the membrane to the other; the loaded transfer step can be isolated from net transport by monitoring protein-facilitated exchange of substrate across the membrane. In an exchange experiment, equimolar concentrations of Na + and L-asp are introduced on either side of the membrane, then substrate is exchanged between the external and internal solutions at equilibrium as the substrate binding site stochastically samples both sides of the membrane (24). By spiking the external solution with 3 H-L-asp, exchange is measured by monitoring 3 H-L-asp accumulation in the lumen of the proteoliposome. Despite several efforts to apply this method to Glt Ph -containing proteoliposomes, we could not measure appreciable radiolabel accumulation. However, we overcame this problem using a solute counterflow assay; here, proteoliposomes are loaded with excess unlabeled substrate (1 mM) and diluted into an external solution containing a trace amount of radiolabeled substrate (100 nM). Here, as in exchange mode, the alternating exposure of the substrate binding site to both sides of the membrane will initially result in exchange of unlabeled substrate in the proteoliposome lumen for radiolabeled substrate in the external solution (Fig. 5, inset, steps 1, 2 and 3). In counterflow, however, the relative concentration of radiolabeled and unlabeled substrate in the proteoliposome lumen essentially eliminates the likelihood of the accumulated radiolabeled substrate from rebinding and exiting back into the external solution. At first, this results in uptake of radiolabel by a process identical to exchange. Eventually, however, the accumulation of radiolabel in the proteoliposome lumen peaks and the accumulated substrate effluxes from the proteoliposomes until the outwardly directed substrate gradient is dissipated (25). If the turnover rate of the transporter is slow, as it is for Glt Ph , then substrate exchange can be monitored separately from the efflux portion of counterflow. Under these experimental conditions, we observe robust accumulation of 3 H-L-asp in proteoliposomes containing Glt Ph WT or intact Glt Ph Xa as a function of time (Fig. 5). In contrast, we observed very slow accumulation of 3 H-L-asp in proteoliposomes containing cleaved Glt Ph Xa: ~8-fold lower accumulation after 10 minutes compared to proteoliposomes containing intact Glt Ph Xa (Fig. 5). Note that we do not continue these experiments long enough to observe the falling phase of the counterflow reaction. This experiment isolates substrate binding/unbinding events (Fig. 5, inset, steps 1 and 3) and the transition of the substrate-loaded transport domain (Fig. 5, inset, step 2) from the rest of the transport cycle. Having already demonstrated that 3L4 cleavage causes minimal perturbation of substrate interaction, this result shows that loop scission substantially attenuates the substrate-loaded transfer reaction.

Efficient transition between substrate-loaded OFS and IFS depends on an intact 3L4
The counterflow experiments implicate 3L4 in the loaded transfer reaction but provide no information on whether 3L4 scission affects the equivalent transition of the apo, substrate-free transport domain (Fig. 5, inset, step 4). We therefore sought to compare the effects of loop scission on the transfer of the empty transport domain, with those of the cleavage on the fully loaded form. Since transport measurements will not report on isolated transitions of substratefree forms of the protein, we measured the effects of loop cleavage on the empty (Fig. 6C, left panel, step 4) and substrate-loaded (Fig. 6C, right panel, step 2) equilibria separately by monitoring the rate of IFS crosslink formation in the absence or presence of saturating substrate concentration, respectively. If the energetics of the transfer reactions are affected then the population of the crosslink-competent, inwardfacing state will be altered and will be reflected as a change in the crosslinking rate. We quantified the rate of IFS crosslink formation of cleaved and intact Glt Ph Xa in the presence and absence of substrate using SDS-PAGE and densitometric analysis ( Fig. 6A and B). IFS crosslink formation was determined for both intact and cleaved protein by quantifying the intensity of the crosslinked band (labeled "CL" in Fig. 6A). For intact protein, the crosslinked band is reflected in a shift of the full length protein band toward faster electrophoretic migration; for cleaved protein, a single crosslinked band is formed by the crosslinking of the 2 proteolytic fragments (e.g. Fig. 1C, lane  7). In the absence of substrate, we observed rapid crosslink formation for both intact and cleaved protein. Cleaved protein crosslinks with a slightly faster rate than intact protein (rate constants of 1.4 and 0.78 min -1 , respectively) although it reaches a lower absolute level of crosslinking (75% compared to ~100% for intact protein, Fig. 6A and B, left panel). Under these conditions, we predominantly monitored crosslink formation of empty transporter in equilibrium between the OFS and IFS (Fig. 6C,  left panel, black box). These results suggest that cleaving 3L4 has little effect on the equilibrium of empty Glt Ph between IFS and OFS (the apo transfer reaction, Fig. 6C, left panel, step 4). This observation is strikingly different from the crosslinking behavior in the presence of saturating concentrations of both substrates. Here, though the rate of IFS crosslink formation for intact Glt Ph was similar to the crosslinking rate without substrate (0.81 min -1 ), the crosslinking is slowed to an immeasurable rate in the cleaved protein, with no more than 3% crosslinking over the 10-minute course of the experiment. (Fig. 6A and B, right panel). These results indicate that cleaving 3L4 results in lower occupancy in the crosslink-competent, inward-facing conformation when in the presence of substrate than in the apo form of the protein. The greatly diminished crosslinking rate for cleaved Glt Ph demonstrates that loop scission specifically impedes the substrate-loaded transport domain from accessing the crosslinkcompetent IFS (Fig. 6C, right panel, step 2). Curiously, during this experiment we observed that crosslinking of intact protein in the presence of substrate (Fig. 6B, left panel, closed squares) does not reach completion. This result suggests that the presence of substrate prevents a population of the Glt Ph (40%) from accessing the inward-facing, crosslink competent state; an unexpected result considering the irreversible nature of the crosslinking reaction. An alternative hypothesis is that the presence of substrate may make one or both of the cysteine residues more accessible to reaction with the crosslinking reagent. If this is the case, and if the rate of this reaction exceeds the rate of entry into the crosslink-competent state (OFS-IFS transition), each cysteine (C55 and C364) may react with different MTS-1-MTS molecules, thus preventing crosslink formation between the two cysteines via a single MTS-1-MTS molecule. We investigated whether this was the case by measuring crosslinking rates under conditions that would slow down the rate of the MTS reaction while leaving the rate of conformational change unaffected. Two parameters were varied that might result in the separation of these two events; concentration of MTS-1-MTS, and the pH of the reaction. In the absence of substrate, decreasing the concentration of crosslinker results in decreased rate of crosslink formation, but the absolute amount of crosslink formation remains essentially the same (Fig. 7A). In the presence of substrate, decreased crosslinker concentration also results in a lower crosslinking rate; however, the absolute amount of crosslinking is greater when lower concentrations are used (Fig.  7B). This demonstrates that when the crosslinking rate is slower, more crosslinking can occur, which we interpret to be due to the protein accessing the crosslink-competent state before the cysteines react with one MTS-1-MTS each. MTS reagents react predominantly with the ionized thiolate form of cysteine which is more abundant at higher pHs, therefore decreasing the pH of the reaction should decrease the rate of crosslinking. Conveniently, the transport rate of Glt Ph remains constant over a wide range of pHs indicating that transport, and any associated conformational shift, is insensitive to pH change (11). Crosslinking timecourses were performed in the presence and absence of substrate at pH 5, 7 or 9. In the absence of substrate, decreasing the pH of the reaction decreased the rate of crosslinking, however the absolute amount of crosslinking was the same regardless of the pH (data not shown). In the presence of substrate, decreasing the pH resulted in a lower crosslinking rate, but resulted in a higher absolute amount of crosslinking compared to the crosslinking performed at higher pHs (data not shown). Together, these findings strongly suggest that the incomplete crosslinking we observe in our experiments is due to reactions of the two cysteines with separate MTS-1-MTS molecules. This phenomenon results in competition between the reaction of interest, crosslinking the cysteines together with one MTS-1-MTS molecule, and a side reaction, where each exposed cysteine reacts with a separate MTS-1-MTS molecule, preventing crosslink formation. However, the crosslinking rates themselves, which are the measurement of interest, are not compromised by the side reaction.

DISCUSSION
Here, we explored the mechanism by which the 3L4 influences the transport cycle in Glt Ph, using a combination of 3L4 proteolytic cleavage, chemical crosslinking, and transport measurements to specifically probe major components of the transport cycle, distinguishing substrate binding, loaded transfer, and apo transfer reactions. In the context of a simplified kinetic scheme, which nonetheless captures the essential features of transport, we find strong evidence that the basic binding/unbinding reactions with substrate and Na + are minimally affected by cutting 3L4, and, surprisingly, that the transfer of the empty transport domain is also preserved. Indeed, our experiments point to the loaded transfer, which includes opening and closing steps and the major translocation of the fully loaded and occluded transport domain across the membrane as the locus affected by cutting 3L4, primarily through effects on the activation energy of translocation. This finding implies that the structure of the unloaded transport domain is significantly different from that of the loaded domain, an idea strongly supported by a recent structure of an apo form of a Glt Ph homolog (19). The simplified model we have focused on to this point (Fig. 1B) distinguishes steps that can be isolated using our functional experiments. Clearly though, each of these reactions comprises multiple substeps which we cannot identify functionally. By lumping the binding reactions of L-asp and Na + our kinetic scheme captures the general features of the binding portions of the transport cycle without requiring a specific binding order. Clearly, a more detailed model would include separate binding steps for L-asp and Na + , ideally in a particular order (26), see Fig. 8 (beige reactions). But by working at zero or saturating substrate/Na + we can safely ignore these details. As we have discussed, the transfer reaction must include multiple substates (8) as shown in the blue reactions in Fig. 8. The transfer steps in our model (Fig. 8, steps 2 and 4) each include three substeps: between outward-open and outwardoccluded, between outward-occluded and inward-occluded (the piston movement), and between inward-occluded and inward-open as indicated in Fig. 8. At the level of our experiments, though, these detailed reactions would be difficult or impossible to isolate; our results do not distinguish which one (or more) of these substeps is affected by 3L4 cleavage in the substrate-loaded form of the transport domain. Our findings do, importantly, clearly and definitively distinguish between the behavior of the apo transport domain and the loaded domain for the overall transition that captures these three experimentally indistinguishable substeps, with only the loaded domain affected by 3L4 cleavage (Fig. 8, blue shading). Due to the position of 3L4, directly overlying the HP2 hairpin, an attractive proposition for its role is to stabilize one or more states of the hairpin, thereby modulating the binding of protein to substrate. Since opening of HP2, as observed in the L-asp-and TBOA-bound structures, may bring it into more intimate contact with 3L4, we considered the possibility that 3L4 could be stabilizing that state, which would in turn affect the binding sites of both substrate and inhibitor. However, we eliminated this possibility after measuring the kinetic parameters for cleaved and intact Glt Ph (Fig. 3). Measurements of both K m and K D suggest that 3L4 cleavage has minimal effects on the substrate binding rates and affinities. Our results therefore focus attention on the kinetic properties of the transfer reactions, showing that the activation energy barrier for one part of this reaction is raised by 3L4 cleavage. The strong temperature dependence of active transport reflects all the temperature dependent reactions that occur during the transport process, including substrate binding and conformational changes. In the Glt Ph homolog EAAC1, the binding of substrate is associated with a low energy of activation and is diffusion controlled; binding of substrate and any associated conformational changes do not significantly contribute to the temperature dependence of transport (27). Additionally, the K m of wild type Glt Ph for Na + is stable between 30 o C and 40 o C despite a large difference in the maximum transport rate (11). Together, these results suggest, at least for Na + , that the substrate binding event for Glt Ph is relatively temperature independent and support the conclusion that the difference in the temperature dependence between cleaved and intact Glt Ph Xa primarily reflect the isolated transfer reactions. What is the structural basis of transport modulation by 3L4? Cysteine-scanning mutagenesis of 3L4 revealed that many single amino acid substitutions result in altered transport activities (17), hinting that specific interactions between residues in 3L4 and another part of the protein are important for transport. Results reported here suggest a model in which the intact 3L4 interacts specifically with the loaded transport domain, thereby stabilizing a specific transition state in the translocation process (Fig. 8), a model supported by the protection of the loop from proteolysis in the presence of substrate (20). We speculate, based on the location of 3L4, that this stabilization might be mediated by interactions between 3L4 and HP2, which directly underlies the loop in the OFS structure. In order for 3L4 to discriminate between substrate-loaded and empty transport domain, as is demonstrated by our data, the transport domain must have an appreciably different form in the presence and absence of substrate. This hypothesis is supported by a newly published structure of a Glt Ph homolog in the apo form which shows subtle but clear changes in the disposition of HP2 (19). This implies a kind of asymmetry in the transport cycle, with the structure of the translocating unloaded domain detectably (to 3L4) different from that of the loaded domain, a possibility not captured in previous models of transport, which assume similar "closed" forms of the protein for both translocation steps (e.g. steps 2 and 4 in Fig. 1B). These observations offer the possibility of additional protein conformations in these transporters that might serve as targets for statespecific, or subtype specific drugs to modulate their function.       , and 3L4 (black line). The simple model used to guide experimental interpretation is shown in pale colors (numbered as in previous figures), while a more detailed model, which includes essential reaction steps, is highlighted. Binding and unbinding of Na + (green circles) and substrate (magenta circle) occur in ordered steps in both inward and outward facing transporters (horizontal, beige reactions). '+' signs indicate that one or two Na + ions may bind in the indicated steps. Elements of the transfer reactions (vertical, blue reactions) include closing, translocation, and opening steps. Our results demonstrate that one or more reactions of the fully loaded transport domain (shaded blue) are affected by 3L4 cleavage suggesting that Na + and substrate binding induce conformational changes in the transport domain and