Identification of a Second Substrate-binding Site in Solute-Sodium Symporters*

Background: Although the solute-sodium symporter (SSS) vSGLT and the neurotransmitter-sodium symporter (NSS) LeuT have similar structural folds, their crystallographically identified substrate sites diverge in location and composition. Results: We identified second substrate sites in two SSSs that align with the crystallographically identified site in LeuT. Conclusion: Substrate transport by SSSs involves two substrate sites. Significance: NSS and SSS share common mechanistic features. The structure of the sodium/galactose transporter (vSGLT), a solute-sodium symporter (SSS) from Vibrio parahaemolyticus, shares a common structural fold with LeuT of the neurotransmitter-sodium symporter family. Structural alignments between LeuT and vSGLT reveal that the crystallographically identified galactose-binding site in vSGLT is located in a more extracellular location relative to the central substrate-binding site (S1) in LeuT. Our computational analyses suggest the existence of an additional galactose-binding site in vSGLT that aligns to the S1 site of LeuT. Radiolabeled galactose saturation binding experiments indicate that, like LeuT, vSGLT can simultaneously bind two substrate molecules under equilibrium conditions. Mutating key residues in the individual substrate-binding sites reduced the molar substrate-to-protein binding stoichiometry to ∼1. In addition, the related and more experimentally tractable SSS member PutP (the Na+/proline transporter) also exhibits a binding stoichiometry of 2. Targeting residues in the proposed sites with mutations results in the reduction of the binding stoichiometry and is accompanied by severely impaired translocation of proline. Our data suggest that substrate transport by SSS members requires both substrate-binding sites, thereby implying that SSSs and neurotransmitter-sodium symporters share common mechanistic elements in substrate transport.


Structure-Structure Alignment between LeuT and vSGLT-
The structure-structure alignment of LeuT (PDB code 2A65) (1) and vSGLT (PDB code 3DH4) (4) was performed using Matt (32) with default parameters. The segments considered in the alignment are the two conserved 5-TM inverted repeats, TM1-10 in LeuT and TM1Ј-10Ј in vSGLT (Fig. 1). The resulting sequence alignment was further manually refined by removing the gaps in helical regions and by local adjustments to align the positions in the two structures that face the permeation pathways. The final alignment is shown in Fig. 1. According to this alignment, the superposition of the structures shown in Fig. 2 is based on the middle portions of the more conserved TM-1/1Ј, -3/3Ј, -6/6Ј, and -8/8Ј, which do not have drastic rearrangements between outward-facing and inward-facing conformations, specifically residues 23-26, 106 -111, 251-254, and 351-358 of LeuT versus residues 65-68, 140 -145, 261-264, and 361-368 of vSGLT.
Induced-fit Docking (IFD) of Substrate to the S e Site-By using the IFD protocol (33), in the presence of a substrate galactose bound in the crystallographically identified substrate site (S e site), a second galactose was docked in the putative S c cavity (see under "Results") of our vSGLT model, which was described previously (20). Specifically, we targeted four sub-pockets as follows: (i) immediately below the gate residue, Tyr-263; (ii) between Tyr-263 and Ser-368; (iii) near the Na ϩ site identified in the vSGLT structure; and (iv) at the midpoint of i and iii. The docking poses from individual trials are pooled together for the clustering analysis. Three representative poses with the highest IFD scores were selected for extensive MD simulations.
Similarly, starting from an equilibrated PutP model based on the crystal structure of vSGLT (PDB code 3DH4) and with a substrate proline bound in the S e site (20), we docked a second proline in its S c cavity, according to the equilibrated poses of the S c substrate in vSGLT (see "Results"). Three representative FIGURE 1. Sequence alignment of vSGLT, PutP, and LeuT. The alignment is based on a structure-structure alignment using Matt (32), followed by manual adjustments to remove gaps in helical regions. TM boundaries for LeuT and vSGLT are defined according to those of Yamshita et al. (1) and Faham et al. (4), respectively. The unwound regions in the TM1/1Ј and TM6/6Ј are underlined. Substrate-binding site residues identified in the crystal structures of LeuT and vSGLT (PDB codes 2A65 and 3DH4, respectively) are highlighted with cyan and red, respectively; the S1 residues of LeuT are from Ref. 52, and the S e residues of vSGLT are defined as having closest heavy atom distance Ͻ4.5 Å to the substrate galactose.
poses with the highest IFD scores were selected for the following MD simulations.

Construction of the Simulation Systems and MD Simulations-
The vSGLT models with the additional S c substrate were reinserted back to the equilibrated explicit water-bilayer-water environment by aligning them to the reference vSGLT model in the original simulation system (20). In this system, the water phase includes Na ϩ and Cl Ϫ ions corresponding to a concentration of 150 mM NaCl; the entire system includes ϳ73,400 atoms, including 171 palmitoyloleoylphosphatidylcholine molecules and ϳ14,000 waters.
Three MD simulations with different initial poses of the S c substrate were carried out for both vSGLT and PutP (Table 1), using Desmond (version 3.0) (34) with the OPLS2005 (35) allatom force field. Periodic boundary condition was employed. The relaxation protocol modified from that developed by Schrödinger Inc. (Desmond 3.0; Schrodinger, Inc., New York, NY), including minimization, heating, and solvation of buried cavities, and equilibrations were carried out before the produc-tion runs. First, two steepest descent minimizations with and without harmonic constraints of 50 kcal/mol on the solute heavy atoms with a maximum of 2000 steps were performed, followed by a 60-ps MD simulation raising the temperature from 10 to 310 K in the canonical ensemble (NVT), during which the solute heavy atoms were restrained. The system was then equilibrated in two successive steps with a Martyna-Tobias-Klein isothermal-isobaric ensemble (NPT) (36), where a harmonic constraint with a force constant of 10 kcal/mol on the solute heavy atoms was applied at the first step and removed afterward. In the production runs, the time step was 1 fs, and a constant surface tension of 4000 bar⅐Å was added (NP␥T). Electrostatic interactions were calculated with particle mesh Ewald (PME) (37) throughout the simulations, and the short range cutoff was set to 10.0 Å.
Clustering of the Substrate Binding Poses in the S c Cavity and MM/GBSA Calculations-After each MD frame was superimposed on the crystal structure (PDB code 3DH4) based on the putative S c site residues (Table 2), the substrate poses in the S c FIGURE 2. Crystallographically identified substrate-binding site of vSGLT is located more extracellularly than that of LeuT. A and B show the superimposed orientations of a LeuT model bound with two substrates Leu (24) and the vSGLT structure (PDB code 3DH4), respectively, based on a structure-structure alignment (see text). The TMs directly involved in substrate binding in either LeuT or vSGLT are in non-white colors and in the same color codes for both panels. In particular, the equivalent TMs in the two 5-TM inverted repeats of LeuT structural fold, e.g. TM1 and TM6, are in the same color. The substrate-binding site in the vSGLT structure, termed S e here, is enclosed by TM1Ј, -2Ј, -6Ј, -7Ј, and -10Ј, whereas for LeuT the crystallographically identified S1 site is formed by residues from TM1, -3, -6, and -8, and the computationally identified S2 site is formed by residues from TM 1, -3, -10, and EL4. C and D are the zoom-in views of the substrate-binding sites in LeuT and vSGLT, respectively. The cavity below Tyr-263 of vSGLT may potentially form a substrate-binding site. Thus Phe-253 of LeuT and Tyr-263 of vSGLT may play similar roles as the gating residues to separate more centrally located substrate-binding sites from the extracellular milieu. The C␣-C␤ bonds of substrate binding residues are shown as sticks (cyan for the S1/S c site and green for the S2/S e sites) with the C␤ atoms represented as small spheres. D, residues Ser-365, Ser-368, and Ser-372 of vSGLT are aligned to Thr-341, Cys-344, and Val-348 of PutP, which have been found to be critical for proline uptake (42). cavity were clustered according to the their center-of-mass (COM) positions with respect to the bound Na ϩ in the crystal structure of vSGLT, in a spherical coordinate system (see the definition in Fig. 3). As the values of the poses vary in a narrow range (within a range of 30°), we clustered the poses according to the distribution of the r and values, and we considered a densely populated region as a cluster (Fig. 4, A and C). For each frame, the MM/GBSA receptor-ligand energy was calculated using Prime (version 3.1; Schrödinger).
Protein Expression and Purification-vSGLT mutations were performed using the QuikChange methodology (Agilent Technologies) with the vSGLT gene in pJexpress as template. All mutants were verified by DNA sequencing, and vSGLT variants were expressed in Escherichia coli XL1-Blue cells. Briefly, an overnight starter culture in Terrific Broth supplemented with 30 g/ml kanamycin was inoculated into 1-liter flasks of the same media to a final absorbance at 600 nm of 0.05 and cultivated aerobically at 37°C. Protein expression was induced at an absorbance at 600 nm of 1.8 with 0.75 M isopropyl 1-thio-␤-D-galactopyranoside for 3 h at 33°C. Cells were collected at 7000 ϫ g for 10 min and frozen at Ϫ20°C prior to the preparation of membrane vesicles with an EmulsiFlex-C3. Membrane vesicles were solubilized by the addition of 2% n-dodecyl ␤-Dmaltopyranoside (Anatrace). vSGLT was bound to a nickel-nitrilotriacetic acid column (GE Healthcare) in 70 mM Tris-Cl, pH 8.0, 150 mM NaCl, 4 mM Na 3 -citrate, 20 mM imidazole, 6% glycerol, 5 mM 2-mercaptoethanol, 0.0174% n-dodecyl ␤-Dmaltopyranoside and eluted in the same buffer with 200 mM imidazole. Peak fractions were concentrated using an Amicon Ultra 50 kDa (Millipore) and loaded onto an analytical Superdex 200 column (GE Healthcare) equilibrated in 25 mM Tris-Cl, pH 7.4, 150 mM NaCl, 2 mM 2-mercaptoethanol, 0.0174% n-dodecyl ␤-D-maltopyranoside. Only the center of the elution peak consisting of 0.5-1 mg of total protein was pooled and used for the equilibrium dialysis assays.
Recombinant PutP wild-type (WT) and variants (containing a C-terminal His 6 tag) (38) were produced in E. coli WG170 and purified as described (39). Mutations of Tyr-248 to Gly, Pro-252 to Cys, Cys-344 to Ala, and Leu-398 to Ser were performed with PCR-based site-directed mutagenesis and confirmed by sequencing. Protein was assayed with the Amido Black method (40).
Binding Studies-Equilibrium dialysis was performed (29) with the HTD96b dialysis 96-well apparatus and 12,000 -14,000 molecular weight cutoff membranes in binding buffer (150 mM Tris/Mes, pH 7.5, 50 mM NaCl, 1 mM tris(2-carboxyethyl)phosphine, 20% glycerol, 0.1% (w/v) n-dodecyl ␤,D- H]proline (90 Ci/mmol; American Radiolabeled Chemicals) was used at the indicated concentration range. 20-l samples were taken from each dialysis compartment separated by the dialysis membrane (total volume per compartment was 50 l) and incubated with scintillation mixture for 24 h prior to photo-multiplier tube-based decays/min) counting. To capture protein bound to the dialysis membrane, the exposed membrane section was excised and subjected to scintillation counting. Nonspecific radiotracer binding to the membrane was determined by performing the assay in the absence of protein, and data were corrected accordingly. Decays/min were transformed into picomoles using known amounts of [ 3 H]galactose or [ 3 H]proline. Scintillation proximity assay (SPA)-based binding of [ 3 H]proline to purified PutP variants was performed as described (39) with 0.92 pmol of purified protein per assay in binding buffer.
[ 3 H]Proline Transport Measurements-Na ϩ -coupled [ 3 H]proline uptake into intact E. coli WG170 harboring plasmids coding for given PutP variants or a control plasmid was performed in 50 mM Tris/Mes, pH 6.0, 50 mM NaCl as described (31). Facilitated diffusion down the proline concentration gradient was performed either in the same buffer or in 100 mM Tris/Mes, pH 6.0 (Na ϩ -free condition) with de-energized WG170 harboring indicated PutP variants. Briefly, 200-l cell suspensions at a protein concentration of 0.07 mg/ml were incubated for 5 min prior to the start of the uptake reaction in the presence of 5 M carbonyl cyanide m-chlorophenylhydrazone (CCCP) and 5 M monensin to dissipate the transmembrane electrochemical Na ϩ and H ϩ gradients (31) prior to the addition of 20 M   (53), which have the z axis perpendicular to the membrane, and the x axis largely pointing from Na2 to the side chain of Tyr-263. The projection of the COM on the xy plane is shown as the point COM xy . "r" is the distance between Na2 and COM; "" is the angle between the x axis and the line connecting Na2-COM xy , and "" is the angle between the z axis and the line connecting Na2-COM.
Data Analysis-Data fits of kinetic analyses were performed using nonlinear regression algorithms in Prism (GraphPad), and errors represent the standard error of the fit.

RESULTS
Crystallographically Identified Substrate-binding Sites in vSGLT and LeuT Are At Different Locations-To identify common and unique key structural elements of LeuT and vSGLT, we performed a structure-structure alignment between the crystal structures of LeuT and vSGLT with Matt (32). When we superimposed LeuT and vSGLT with the core TM regions (TM1-10 of LeuT and TM1Ј-10Ј of vSGLT; Fig. 1), we found that the substrate-binding site of vSGLT, which is formed by residues from TM1Ј, -2Ј, -6Ј, -7Ј, and -10Ј, is situated more extracellularly than the central occluded substrate (S1) site in LeuT, which is enclosed by TM1, -3, -6, and -8 (Fig. 2). Based on the superposition shown in Fig. 2, the bound substrates, galactose in vSGLT and leucine in LeuT, have a COM distance of 8.0 Å. In fact, the vSGLT substrate-binding site is at a location between the S1 site and the extracellular second substrate (S2) site of LeuT, the latter of which was identified using computational modeling and simulations, as well as radiotracer and flux measurements (24). Whereas the S2 site in LeuT is located in the middle of the so-called extracellular vestibule (EV) that is composed of residues shown to participate in binding tricyclic antidepressants (15) and selective serotonin reuptake inhibitors (41) in outward-facing conformations, the substrate-binding site of vSGLT is at the intracellular end of the collapsed EV in the inward-facing conformation. To account for these differences, we refer to the substrate-binding site revealed by the crystal structure of vSGLT (PDB code 3DH4) as the extracellular substrate (S e ) site.
In particular, Tyr-263 from TM6Ј of vSGLT at the intracellular end of the galactose-binding site is similarly located as Phe-253 from TM6 of LeuT, although there are subtle divergences between the unwound regions of TM6Ј in vSGLT and TM6 in LeuT in terms of length and orientation; Tyr-263 of vSGLT and Phe-253 of LeuT are both at the bottom of the EV and above the cavity formed by TM1Ј/1, -3Ј/3, -6Ј/6, and -8Ј/8, i.e. the S1 site in LeuT. Thus, although the vSGLT substrate galactose is located right above (extracellular to) Tyr-263, the S1 substrate of LeuT is positioned right below (intracellular to) Phe-253, which was found to act as the gate at the extracellular side of the S1 site; it separates the S1 site from the EV (1, 24). On the extracellular side of the S e site of vSGLT, Phe-424 and Gln-428 from TM10Ј can be aligned precisely to Leu-400 and Asp-404 of LeuT, residues that have been found to be involved in S2 substrate binding (24 -26).
Our structure-structure alignment of vSGLT and LeuT also aligns the deduced Na ϩ -binding site of vSGLT (4) precisely to the Na2 site of LeuT (1). Curiously, near this Na ϩ -binding site and below Tyr-263 of vSGLT, there is an empty pocket formed by TM1Ј, -3Ј, -6Ј, and -8Ј, which are aligned to the TMs that enclose the S1 site of LeuT. In the inward-facing conformation of vSGLT, this pocket is highly exposed to the intracellular water phase (3,4). Although the carboxyl group of the S1 substrate in LeuT directly coordinates the bound Na ϩ in the Na1 site and is in close vicinity to Na2 (with a COM distance of ϳ6.5 Å), the substrate bound in the S e site in vSGLT is significantly further away from the crystallographically identified single Na ϩ site (with a COM distance of ϳ14.5 Å). Thus, we hypothesize that in another conformational state the pocket below Tyr-263 in vSGLT may form a central substrate-binding site that is close to that Na ϩ site. In the following, we termed this pocket of vSGLT as the central substrate (S c ) site, which structurally aligns with the S1 site of LeuT.
A recent study of PutP using the substituted cysteine-accessibility method (SCAM) supports the presence of the S c site (42). In this study, labeling of the individual thiol group at residue positions 344, 347, 348, and 351 was reduced in the presence of the substrate proline but not by Na ϩ alone (42). Interestingly, the aligned residues at these positions in vSGLT face the S c site, but are not involved in forming the crystallographically identified S e site in vSGLT (Fig. 2D).
Potential Substrate-binding Residues in the S c Site of vSGLT Align to the S1 Residues of LeuT-To specifically identify the residues in the S c site that are critical for substrate binding, using the comprehensive vSGLT model constructed and equilibrated previously (20), we first performed induced-fit docking to position a second substrate galactose in the S c site to identify its optimal poses in the presence of the galactose bound in the S e site.
The resulting poses of galactose in the S c site were clustered into three groups (Table 1). A representative S c pose from each group was selected based on the MM/GASA binding energy (see "Experimental Procedures"). After reinsertion of these vSGLT models bound with galactose at both the S c and S e sites back into the explicit water-membrane simulation system, they were further relaxed and characterized with MD simulations. The MD simulations were carried out until the root mean square deviations of the TM regions plateaued for at least 100 ns (Table 1). Whereas the S e substrate stayed in the pose identified in the crystal structure in all simulations, the substrate in the S c site fluctuated around the original docked location but remained within this site. We clustered the S c substrate poses from all MD trajectories into three clusters (Fig. 4A). Strikingly, for the two clusters with more favored binding energies (clusters 2 and 3 in Fig. 4, B and E), most of the substrate-interacting residues are aligned to the S1 residues of LeuT (Table 2). Specifically, in the MD frame with the most favored galactose binding energy, among the 10 galactose-interacting residues, seven can be aligned to the S1 substrate-binding residues in LeuT. In this frame, the hydroxyl groups of S c galactose appear to optimally form H-bonds with the side chains of Ser-66, Arg-273, Ser-365, and Ser-368, and the backbone of Tyr-269 (Fig. 5). Interestingly, the aligned Na2 binding residues Ser-365 in vSGLT and Ser-355 in LeuT are similarly able to both coordinate the Na ϩ ion and be in close proximity to the substrate. In contrast, among the residues forming the S e site, only the gating residues Asn-64 and Tyr-263 and the nearby Trp-264 can be aligned to the S1 residues of LeuT (Table 2).
vSGLT Has a Molar Binding Stoichiometry of 2-To experimentally test the computational findings, we performed [ 3 H]galactose saturation binding using equilibrium dialysis with known amounts of purified vSGLT to deduce the galactose-to-vSGLT molar binding stoichiometry (MBS). This method allowed for the assessment of the binding activity without relying on a potential bias introduced by the calculations of counting efficiency in different detection modes, as all samples and calibration standards were subjected to the same scintillation counting format (29). The MBS of [ 3 H]galactose binding to vSGLT in the presence of 50 mM Na ϩ was concentration-dependent and reached a maximum of ϳ2 (Fig. 6A). Subjecting the data of three independent experiments to a single-site global fitting analysis (GraphPad Prism 5) yielded a K d (concen-

Composition of the S c and S e binding sites of vSGLT and PutP
The binding site residues in vSGLT and PutP are based on the clusters with averagely more favored MM/GBSA binding energy, i.e., clusters 2 and 3 for vSGLT and clusters 1 and 3 for PutP. Residues within 4.5 Å of the ligand in Ͼ50% of the frames in each cluster are listed here.

Second Substrate-binding Site in the SSS Family
tration at half-maximum binding) of 179.2 Ϯ 19.9 M and a B max of 2.07 Ϯ 0.07 molecules of galactose bound per molecule of vSGLT.
Mutations of the S c or S e Site Residues of vSGLT Reduce the MBS to One-To assess the impact of disrupting the substrate binding in the S e site, we targeted Phe-424 and Glu-88 that are on the extracellular side of the S e site and relatively far from the proposed S c site. Specifically, Phe-424 aligns with Leu-400 in LeuT, a residue that has been shown to participate in the binding of substrate in the S2 site of LeuT (24), whereas Glu-88 has been previously shown to be important for galactose transport (4). By mutating Phe-424 to Ser or Glu-88 to Ala, the MBS of two observed for vSGLT-WT was reduced to ϳ1. Fitting the data to a one-site model showed small alterations of the K d values (133.3 Ϯ 12.1 M and 176.5 Ϯ 20.3 M for E88A and F424S), whereas the MBSs were reduced to 1.20 Ϯ 0.03 and 1.20 Ϯ 0.04 for E88A (Fig. 6B) and F424S (Fig. 6C), respectively.
To disrupt the proposed S c site, we substituted Ser-368, an S c residue relatively far from the S e site and the Na ϩ site, with Ala.  (Fig. 6D). Taken together, the saturation binding data support our hypothesis that under equilibrium conditions vSGLT, like LeuT, can simultaneously bind two substrate molecules, only one of which has been identified in the available crystal structures of vSGLT (4).
Homologous PutP Can Bind Two Substrate Molecules-To test whether other SSS members also exhibit an MBS larger than unity, we performed equilibrium dialysis-based saturation binding of [ 3 H]proline to purified PutP, a well characterized and experimentally tractable SSS (31,38,(43)(44)(45)(46), which has a sequence homology of 42% (sequence identity of 20%) to vSGLT in the TM1Ј-10Ј (Fig. 1). Consistent with the simultaneous occupation of PutP by two substrate molecules under equilibrium conditions, the proline-to-PutP MBS was ϳ2 (Fig.  7A). However, a plateau in the binding curve (between 5 and 15 M [ 3 H]proline) was noticeable for PutP (Fig. 7A, inset), which prevented reliable fitting with one-site or two-site models. Therefore, the observed two phases were fit independently with one-site models. Fitting the data ranging between 0 and 15 M  (Fig. 7A).
In addition, we performed proline binding studies with PutP by means of the scintillation proximity assay (SPA), a rapid and sensitive method to assess the detailed binding kinetics (Fig. 7, B-F) (24,39). Consistent with the equilibrium binding data (Fig. 7A) Mutations of the S c or S e Residues in PutP Reduce the Prolineto-PutP MBS-To gain insight into the determinants of proline binding by PutP at the atomistic level, based on the fact that PutP shares significant sequence homology with vSGLT and can similarly bind two substrate molecules simultaneously, we carried out computational analyses with vSGLT-based PutP homology models. Previously, we built and equilibrated a PutP model with a proline bound in the S e site (20), the equilibrated  state of which has a root mean square deviation of ϳ2 Å to the vSGLT structure. As proline is smaller than galactose, we found the S e site of PutP is formed by residues from TM1Ј, -2Ј, -6Ј, and -10Ј but does not include any TM7Ј residue ( Table 2). Starting from this model, we docked a second proline in the sub-pockets of the S c site that align to those in vSGLT (Table 1), and we selected three representative S c poses for the following MD simulations to explore the potential S c binding residues. We carried out clustering analysis of the resulting S c proline binding poses from the MD simulations, and we calculated the MM/GBSA energies to identify the poses with more favored binding energies (Fig. 4, C, D, and F). Similar to the situation in vSGLT, we found many S c residues of PutP from TM1Ј, -3Ј, -6Ј, and -8Ј are aligned to the S1 residues in LeuT ( Table 2).
To experimentally validate the computationally identified substrate-binding modes in PutP, we created mutants of the residues located at the intracellular end of the proposed S c site and the extracellular end of the S e site, i.e. mutating Cys-344 to Ala and Leu-398 to Ser, respectively. These residues are aligned to Ser-368 and Phe-424 of vSGLT (see above). Furthermore, we also mutated Tyr-248, a residue located between the S e and S c sites, which can potentially interact with the bound substrates from both sites (Table 2), and Pro-252, another residue facing the proposed S c site and located more intracellularly than Tyr-248 in TM6Ј. The binding data observed for PutP-L398S was best fit by a single-site model with a K d of 6.1 Ϯ 2.1 M and an MBS of 1.0 Ϯ 0.1 proline-to-PutP (Fig. 7C). Proline binding by PutP-C344A resulted in a complex binding isotherm that saturated at an MBS of 1.05 Ϯ 0.05 proline-to-PutP and was fit best with two separate single-site models with a K d of 4.57 Ϯ 0.64 M for data points between 0 and 10 M and a K d of 18.3 Ϯ 3.08 M for data points between 15 and 70 M (Fig. 7D). The reduction of the MBS to ϳ1 was also detected when Tyr-248 was replaced with Gly ( Fig. 7E) 7F).

Mutations of Residues in the S c or S e Site in PutP Disrupt
Proline Transport-When we measured the time course of 2 M [ 3 H]proline transport in the presence of saturating NaCl (10 mM) (31), we found the initial transport rates of PutP-L398S and -C344A to be reduced to 13.0 Ϯ 0.2 and 28.8 Ϯ 0.5% that of PutP-WT, whereas their steady-state level of accumulation was 13.3 Ϯ 0.3 and 23.0 Ϯ 0.8%, respectively, of the WT value (Fig.  8A). The initial rate and steady state level of Na ϩ -coupled proline transport by PutP-Y248G were 6.4 Ϯ 0.5 and 10.7 Ϯ 2.1% compared with those of PutP-WT. PutP-P252C exhibited 43.7 Ϯ 2.3 and 71.6 Ϯ 2.7% of the initial rate of transport and steady state level of proline accumulation, respectively, of the rates measured for WT.
The alterations in the time-dependent proline uptake are reflected in the kinetics of active transport. Compared with PutP-WT, all mutants exhibited reduced maximum transport velocities (V max ) in the following order: PutP-WT (24.6 Ϯ 0.2 nmol ϫ mg Ϫ1 ϫ min Ϫ1 ) Ͼ PutP-P252C (52.1 Ϯ 0.5% of the
We then evaluated the impact of the mutations on proline counterflow (Fig. 8G). Here, the rapid dilution of cells preloaded with 1 mM proline into buffer containing 1 M [ 3 H]proline leads to the exchange of radiolabeled and nonlabeled substrate until equilibrium is reached. In similar fashion to Na ϩ -coupled transport, proline counterflow by the mutants exhibited reduced initial rates when compared with PutP-WT (154 Ϯ 17.9 pmol ϫ min Ϫ1 ϫ mg cell protein Ϫ1 ) before reaching equilibrium after about 1 h of incubation as follows: PutP-P252C (51.6 Ϯ 2.1%) Ͼ PutP-Y248G (19.8 Ϯ 0.6%) Ͼ PutP-C344A (14.1 Ϯ 1.2%) Ͼ PutP-L398S (8.5 Ϯ 0.3%). As determined with immunoblotting, the reduced transport activity observed for the protein mutants could not be attributed to different amounts of proteins in the bacterial membrane.
To assess whether the reduced transport activity of the tested PutP mutants could be specifically attributed to disrupted proline but not the coupling of Na ϩ and proline translocation, we measured the facilitated diffusion of 20 M [ 3 H]proline in the presence of the uncouplers monensin and CCCP (Fig. 8H). This established approach (31) allows for the assessment of the uncoupled substrate flux down its concentration gradient after dissipating the electrochemical Na ϩ and H ϩ gradients (⌬ Na ϩ and ⌬ H ϩ). Furthermore, to test whether the mutations affect proline transport through defects in Na ϩ binding, we performed the facilitated diffusion of proline also in the virtual absence of Na ϩ (in presence of 100 mM Tris/Mes, pH 6.0). Consistent with Na ϩ -coupled proline transport and proline counterflow (Fig. 8, A and G), the initial rates of facilitated diffusion of [ 3 H]proline down its concentration gradient by PutP-L398S, PutP-C344A, and PutP-Y248G were reduced by ϳ90% compared with PutP-WT, whereas PutP-P252C showed an intermediate effect (ϳ50% reduction), regardless of whether the assay was performed in the presence or absence of Na ϩ .

DISCUSSION
Among the crystal structures of transporters with the LeuTlike structural fold, the number and location(s) of the substratebinding site(s) in each transporter family are not necessarily identical (Table 3). This raises the question whether such a divergence reflects unique mechanistic features in each family or whether previously unknown commonalities are yet to be revealed, as only certain conformational states for each transporter are readily amenable to crystallography. In this study, we evaluate whether under equilibrium conditions the SSS members vSGLT and PutP can bind substrate in an additional computationally identified site at a location similar to the S1 site of LeuT that is distinct from the galactose binding (S e ) site identified in the crystal structure of vSGLT. Our modeling and simulation results indicate that the pocket below the conserved and functionally important tyrosine residue (Tyr-263 in vSGLT (3,21) and Tyr-248 in PutP (45)) of TM6Ј, which is enclosed by TM1Ј, -3Ј, -6Ј, and -8Ј, may form such an additional substratebinding site, termed the S c site. The majority of the residues forming the S c site of vSGLT as revealed in our simulations are well aligned to the S1 site residues in LeuT (Table 2). However, because the computational study is based on the inward-facing structure of vSGLT, in which the TMs that form the site are unlikely in their optimal positions to accommodate the S c substrate, we expect the residue composition of the S c site to be somewhat different in an inward-occluded state, in which the pocket is prone to binding of the S c substrate. Such a dynamic formation of a binding site in different conformational states along the transport cycle has been observed previously in LeuT and is likely of mechanistic significance; the significant rearrangements of the intracellular portions of TMs from the inward-close to inward-open states of LeuT (PDB code 2A65 and 3TT3, respectively) (1, 18) lead to the release of substrate from the S1 site during the transition to the inward-open state.
Interestingly, recent simulations from Cheng and Bahar indicate that during LeuT's transitions toward the inward-facing state, the S2 substrate can stably bind right above the gating residue Phe-253 (which aligns to the Tyr-263 of vSGLT and Tyr-248 in PutP) with increased affinity (47). It is thus tempting to speculate that the vSGLT structure (PDB code 3DH4) captured a similar intermediate state with the S e site occupied but with the S c site not in its optimal configuration to bind the substrate.
To experimentally validate our computational findings, we performed saturation binding experiments with known amounts of protein to determine the MBS. Consistent with our computational predictions, equilibrium dialysis-based saturation [ 3 H]galactose binding confirmed that one molecule of vSGLT can simultaneously bind two galactose molecules under equilibrium conditions. An MBS of two was also reported for Leu and Ala binding to LeuT by us and others (24,27).
In support of the existence of two substrate sites in vSGLT, individual mutations of residues located in either the S e or the S c sites reduced the MBS of 2 to ϳ1. Obtaining slightly larger than one MBS in those vSGLT mutants may indicate that the mutations did not entirely abolish substrate binding in a particular site but may reflect a severely reduced substrate interaction with the remaining binding residues in that site. From the macroscopic perspective, as the measured stoichiometry is an ensemble average, intermediate stoichiometry may reflect the partial loss

TMs forming the non-ion substrate binding sites in the representative crystal structures of transporters with LeuT-like structural fold
We consider that a TM is involved in forming a binding site when a heavy atom of any of its residues is within 4 Å of the bound substrate. Note, the nomenclature of the transporter conformational states are not consistently used among different groups. We largely classify the transporter conformations into outward-facing, inward-facing, intermediate, or occluded ("closed") states.
of binding to a particular site in a fraction of the protein molecules. Interestingly, in our MD simulations, although the S c and S e sites in vSGLT are separated by the gating residues Tyr-263 and Asn-64, we observed communication between the two substrate-binding sites. Thus, in the presence of S c galactose, binding of S e galactose is weakened (with Asn-64, Trp-264, Phe-424, and Gln-428 interacting less frequently), and S e has a tendency to move upward, which demonstrates the dynamics a binding site undergoes during the transport cycle.
We then extended our studies to PutP, an experimentally tractable SSS model system. Similar to vSGLT, our simulations defined S c -and S e -binding sites in PutP, which are separated by the gating residues Asp-55 and Tyr-248, two residues that have been shown to be functionally important (Table 4) (31,45). In analogy with binding studies in vSGLT, equilibrium dialysisbased [ 3 H]proline binding by purified PutP revealed an MBS of two as well. Fitting the experimental data to two individual single-site models revealed K d values that are different by 1 Functional effect of mutations in the S c and S e sites of SSS proteins a SC followed by a number indicates the substitution of the native residue at the indicated position with a cysteine in a cysteine-free protein. b X represents several replacements. c Boldface type indicates the position studied in this work. Pro, proline; Gal, galactose; NEM, N-ethylmaleimide; F5M, fluorescein 5-maleimide. order of magnitude for two sites. Similar to this observation, binding of Ala to LeuT exhibits a complex curve (24,25) as we observed here for proline binding to PutP. Note that previous binding experiments involving PutP did not exceed radiolabeled proline concentrations of 20 M, thereby missing the second component of the curve (38,48). Mutation of Cys-344 or Leu-398, two residues that are located in the S c and S e sites, respectively, reduced the proline-to-PutP MBS to about one. However, due to the probable complexity of the allosteric interaction of substrate binding to each site, including how binding to one site interferes with the interaction of substrate with the other site, the actual "affinity" of each of the sites may be difficult to determine experimentally.
Implicating the functional importance of two substratebinding sites and their allosteric interaction for substrate transport, the reduction of the MBS from 2 to ϳ1 (observed for PutP-Y248G, PutP-C344A, and PutP-L398S) is accompanied by a severe reduction of the proline translocation. This reduction cannot be attributed to defects in Na ϩ binding, as proline uniport (facilitated diffusion along the proline concentration gradient) in the absence of Na ϩ shows a similar trend observed with Na ϩ -coupled or Na ϩ -uncoupled transport modes. In contrast, differential effects on uniport in the absence or presence of Na ϩ have only been observed for mutants that were shown to disrupt Na ϩ binding directly (31). Replacing Pro-252 with Cys moderately disrupts the binding. Consistently, this mutant shows the smaller effect on all tested transport modes.
Reversibility of active transport has been reported more than 40 years ago (49) and has been generally accepted as a common feature of ion-coupled substrate transport according to which the polarity of transport depends on the direction of the concentration gradient of the transported ions and/or substrates (50). Whereas previous transport measurements of PutP inversely inserted into the membrane of proteoliposomes have revealed its reversibility (51), our counterflow results provide the basis for a detailed analysis of the role of both substrate sites in the reverse transport PutP.
Taken together, our results highlight the functional importance of two substrate-binding sites in vSGLT and PutP, which may be representatives for the other members of the SSS family ( Table 4). The location of the S c site (similar to that of the S1 site in LeuT) brings the substrate and Na ϩ binding into close proximity, an observation with wide ramifications for future comparative studies that aim to identify the mechanistic elements of Na ϩ -coupled substrate transport by SSS that are either unique or common to other transporters with the LeuT-fold.