Determinants of Substrate and Cation Transport in the Human Na⁺/Dicarboxylate Cotransporter NaDC3^*

Template Selection

The Na⁺-dependent dicarboxylate transporter from V. cholerae (vcINDY) is currently the only experimentally determined structure of an SLC13 homolog (13). To confirm the template structure, we used the fold recognition and modeling servers HHpred (16) and I-Tasser (17). In addition, we analyzed the NaDC3 entry in the Transporter Classification Database (18) and the previous analysis of the human solute carriers (19, 20). NaDC3 and vcINDY both belong to the DASS family of TCDB (2.A.47), increasing our confidence in the fold assignment.

NaDC3-INDY Alignment

Initial alignment between NaDC3 and vcINDY (sequence identity of 33%) was obtained using the Promals3D server (21) using the default parameters. The alignment included, as input, the sequences of the human SLC13 members, including NaDC3, and the structure of vcINDY. The initial alignment was refined on the basis of the previously published comprehensive alignment of the SLC13 family with vcINDY (13). Finally, 428 residues of NaDC3 were modeled, covering 75% of the protein sequence. The following four segments distant from the estimated binding site were excluded from modeling: the N terminus, the loop between HP_in and transmembrane helix 5a (TM5a), the loop between TM6 and TM7, and the loop between TM8 and TM 9 (Fig. 1).

Model Construction

We used the “automodel” routine of MODELLER-9 v. 11 to generate the initial 100 NaDC3 models (22). NaDC3 was modeled with the non-protein atoms of succinate on the basis of the coordinates of these atoms in the template structure, (i.e. using the env.io.hetatm = TRUE and the env.io.water = FALSE flags). The models were assessed using Z-DOPE, a normalized atomic distance-dependent statistical potential on the basis of known protein structures (23). The Z-DOPE score of the top model was −0.17, suggesting that ∼55% of the Cα atoms of the model are within 3.5 Å of their correct positions (24).

Side Chain Refinement

We used SCWRL4 to repack side chains on a fixed backbone of key residues of the initial model from MODELLER (25). The coordinates of the sodium ion from the initial model were used as steric constraints for the side chains. SCWRL4 was run on different combinations of residues, including Ser-143 alone, four binding site residues (i.e. Thr-253, Pro-255, Thr-145, Ser-143, and Ser-526), residues located within 6 Å from the coordinates of succinate in the initial model, and all residues. We evaluated the refined models on the basis of visual analysis of the protein-ligand complexes in the context of the experimental functional data.

Molecular Dynamics Simulations

Molecular dynamics simulations were used as a control to verify that our final model (with refined side chains) does not undergo significant side chain rearrangements upon minimization. In particular, we used GROMACS4 molecular dynamics code as described previously (26). The model was subjected to 10,000 steps of conjugate gradient minimization under the Amber99SB-ILDN force field (27, 28). We used two different protocols. First, the sodium ion was removed from the model, and an implicit model for the solvent on the basis of a generalized Born formalism was used to account for the membrane hydrophobic environment. A dielectric constant equal to 2 was used to model the membrane interior. Second, the sodium ion was present in the model, and the system was simulated “in vacuum” where coulomb interactions were screened by using a dielectric constant equal to 2. All bond lengths were constrained to their equilibrium values using the linear constraint solver (LINCS) algorithm (29). A time step of 2 fs was adopted. A cutoff of 1.0 nm was used for the Lennard-Jones and the electrostatic interactions. Finally, to confirm that removing the loops did not have a significant effect on the simulations, we built the missing loops of the model and ran a minimization in an implicit membrane. No significant movement was observed (root mean square deviation of 0.5 Å), increasing our confidence in the model.

Molecular Docking

Docking of NaDC3 substrates was performed with DOCK 3.5.54 (30, 26), which calculates scores of docking poses with van der Waals, Poisson-Boltzmann electrostatic, and substrate desolvation penalty terms. The transporter model with refined side chains was prepared by removing all non-protein atoms except for the sodium ion. Binding site residues were identified as residues with at least one atom within 10 Å of any heavy atoms of the ligand citrate from the initial model using the program FILT (from the DOCK 3.5 distribution) (30).

Expression of NaDC3 Mutants in COS-7 Cells

Eighteen single amino acids from NaDC3 in the pcDNA3.1 vector were mutated individually using the QuikChange site-directed mutagenesis kit (Stratagene) as described previously (31). COS-7 cells (ATCC, catalog no. CRL-1651) were cultured in Dulbecco's modified Eagle's medium containing Glutamax and 25 mm HEPES (Invitrogen) supplemented with 10% heat-inactivated fetal calf serum, 100 units/ml penicillin, and 100 μg/ml streptomycin at 37 °C in 5% CO₂. Cells were plated on collagen-coated 24-well plates at 0.6 × 10⁵ cells/well and transfected with NaDC3 plasmids using FuGENE6 (Roche) at a 9:3 ratio (1.8 μl FuGENE6 and 0.6 μg of plasmid DNA) (31).

Transport Assays

Transport assays were carried out 48 h after transfections, as described previously (31, 32). The sodium buffer contained 140 mm NaCl, 2 mm KCl, 1 mm MgCl₂, 1 mm CaCl₂, 10 mm HEPES, pH adjusted to 7.4 with 1 m Tris. Choline and lithium buffers contained equimolar choline chloride or LiCl, respectively, in place of NaCl. For the assays, each well was washed twice with choline buffer and then incubated with 0.25 ml of sodium buffer containing [¹⁴C]succinate (∼50 mCi/mmol, ∼10 μm, Moravek) for 30 min at room temperature. The uptake assays were stopped, and surface radioactivity was removed with four 1-ml washes of choline buffer. Cells were dissolved in 1% SDS, transferred to scintillation vials, and counted. For all experiments, uptake rates in vector-transfected cells were subtracted from those measured in NaDC3 plasmid-transfected cells to correct for background counts.

In sodium activation experiments, the rate of [¹⁴C]succinate transport was measured in transport buffer containing Na⁺ concentrations between 0–140 mm, with NaCl replaced iso-osmotically by choline chloride. Kinetic constants were calculated by nonlinear regression to the Hill equation, v = V_max × [Na⁺]^nH / (K_Na^nH + [Na⁺]^nH), where v is the initial rate of succinate uptake, V_max is the maximum rate at saturating sodium concentration, [Na⁺] is the sodium concentration, K_Na is the sodium concentration that produces ½ V_max, and n_H is the Hill coefficient.

Dual-label Competitive Transport Experiments

The succinate:glutarate transport specificity ratio (TSR) was determined using dual-label transport assays in which sodium transport buffer containing both 10 μm [³H]succinate and 20 μm [¹⁴C]glutarate was added to the cells in 24-well plates as described previously (33). The transport specificity ratio was calculated using the following: TSR = (ν_succinate / ν_glutarate) × ([glutarate] / [succinate]), where ν_succinate and ν_glutarate are the rates of transport of [³H]succinate and [¹⁴C]glutarate and [glutarate] and [succinate] are the concentrations of glutarate and succinate (34).

Statistics

Duplicate or quadruple measurements were made for each data point. The experiments were repeated with at least three different batches of transfected cells from different passage numbers. Significant differences between groups were identified by Student's t test or analysis of variance with p < 0.05.

NaDC3 Homology Model

The NaDC3 structure was modeled using MODELLER-9 v. 11 (22, 20) on the basis of the x-ray structure of vcINDY, which represents an inward-facing citrate- and Na⁺-bound conformation (Figs. 2 and 3). vcINDY and NaDC3 are members of the DASS family (14), share a sequence identity of ∼35% in their aligned regions, and also have a highly conserved binding site, which increases confidence in template selection and alignment (13) (Fig. 1). The initial model was assessed with Z-DOPE, a normalized, atomic distance-dependent statistical potential on the basis of known protein structures (23). The Z-DOPE score of the top model was −0.17, suggesting that ∼55% of the Cα atoms of the model are within 3.5 Å of their correct positions (23). The final refined NaDC3 model contains the entire transmembrane domain of the protein (i.e. the 11 transmembrane helices), including the residues that make up the substrate- and Na⁺-binding site (Figs. 2 and 3). Particularly, two helical hairpin loops, HP_in and HP_out, are inserted into the membrane from its opposing sides, similar to the corresponding loops in the vcINDY structure (4, 13). Each of those loops contains a conserved motif, the SNT motif (residues 143–145 and 483–485), that is highly conserved among SLC13/DASS transporters and contains functionally important residues (Figs. 1, 3, and 4) (4).

Na⁺ Binding Site

The majority of the interactions between the Na1 sodium ion and NaDC3 are conserved with the vcINDY structure (Figs. 3B and 4). In particular, the oxygen atoms of the Ser-139 and Asn-144 side chains are predicted to make hydrogen bonds with Na⁺, similar to Ser-146 and Asn-151 in vcINDY (Figs. 3 and 4). Asn-144 is part of the N-terminal SNT motif (residues 143–145). Furthermore, the backbone atoms of Gly-252 and Ser-139 are in close proximity (within 4 Å) to Na1, adopting almost identical conformations as the backbone atoms of the corresponding residues in vcINDY (i.e. Gly-199 and Ser-146, respectively). The side chain hydroxyl group of Thr-253 in NaDC3 also coordinates the Na⁺ ion, different from that of the corresponding residue in vcINDY (i.e. Ser-200). However, it is plausible that Thr-253 in NaDC3 and Ser-200 of vcINDY have similar side chain configurations (Fig. 4). The residues that make up the cation binding site are highly conserved among the SLC13 family members, even in other DASS members such as the Drosophila INDY, that do not transport Na⁺ (35).

Interaction with Substrates

Succinate, a four-carbon dicarboxylate, is a natural substrate of both NaDC3 and vcINDY. We docked succinate against the NaDC3 model binding site using DOCK (30) (Fig. 3). The oxygen atom from one of the carboxyl groups of succinate is predicted to be coordinated by the backbone oxygen atom of Ser-143 and the backbone nitrogen atom of Asn-144 of the N-terminal SNT motif as well as by the backbone nitrogen atom of Ala-254 and the side chain hydroxyl group of Thr-253 (Fig. 3). Furthermore, the C_β and C_γ2 atoms of Asn-144 and Thr-145, respectively, face the binding site (e.g. they are both 4.3 Å from the closest succinate carbon), likely contributing to the increased substrate binding affinity of the hydrophobic moieties of the substrates via van der Waals interactions and the hydrophobic effect. Additional binding site residues with hydrophobic groups with the propensity to interact with substrates via similar interactions include Pro-528, Thr-253, and Ala-254 (Figs. 3A and 4).

Two key amino acid substitutions between NaDC3 and vcINDY likely lead to differences in the binding modes of their substrates. Ser-200 and Pro-201 in vcINDY are replaced by Thr-253 and Ala-254 in NaDC3, respectively, resulting in small changes in the size and shape of the region in the binding site close to Na1 (Figs. 3 and 4). For example, the binding site region occupied by the imino ring of Pro-201 in vcINDY becomes accessible to the substrate in NaDC3 upon substitution of this residue with Ala-254. This interaction is very different from the template vcINDY structure, in which citrate does not make similar interactions (Fig. 4).

We docked additional NaDC3 substrates, including α-ketoglutarate, citrate, glutarate, and malate, against the model binding site (Fig. 5). These molecules are larger than succinate and include dicarboxylates with more than two carbon atoms connecting the carboxyl groups (e.g. glutarate) and tricarboxylates (e.g. citrate). The predicted NaDC3-substrate complexes indicate that these larger molecules can fit into the binding site by interacting with Na⁺, similar to succinate, as well as by interacting with additional residues that overlap with the corresponding citrate binding site in the vcINDY structure (Figs. 3, 4, and 5). For example, citrate is predicted to form polar interactions with the side chain hydroxyl group of Thr-485 of the C-terminal SNT motif in NaDC3 (residues 483–485), which is similar to the interaction between citrate and Thr-379 in the vcINDY structure (Figs. 4 and 5D). Thr-485, which is not absolutely conserved among all members of the family, is not predicted to make a similar interaction with all docked substrates (e.g. succinate (Fig. 5C)). Therefore, as visualized by surface representation of the binding site (Fig. 5), both SNT motifs (NaDC3 residues 143–145 and 483–485) are located in the binding site and, likely, have key roles in determining substrate specificity.

Model Testing Using Site-directed Mutagenesis

We selected residues for mutagenesis that were predicted by our model to interact directly with substrate or Na⁺ in the NaDC3 binding site. The residues were Ser-143, Asn-144, and Thr-145 from the N-terminal SNT motif at the tip of HP_in; Thr-253 and Pro-255 from transmembrane helix 5a; Ser-483, Asn-484, and Thr-485 from the C-terminal SNT motif at the tip of HP_out; and Thr-527 in TM10b (Fig. 1). We also mutated residues Gly-164 and Cys-517, which have been shown to be functionally important in previous studies, even though they are not predicted to interact directly with substrates in the model. Gly-164 is located in the intracellular loop between HP_in and transmembrane helix 5a and appears to be important for distinguishing between glutarate and succinate in the mouse (mNaDC1) and rabbit (rbNaDC1) NaDC1 homologs (33). Finally, we mutated Cys-517 in TM10a of NaDC3 because the corresponding residues in NaDC1 (Ser-512 (rbNaDC1) and A504 (mNaDC1)) determine substrate affinity (36).

Mutants were screened for succinate transport activity after transient transfections of COS-7 cells. Mutations of residues predicted in the model to interact with substrate or cations produced decreased succinate transport activity (Fig. 6). Mutants N144A, N484A, and T485P were inactive, as was the double mutant P255T/T485P, made in an effort to rescue T485P. The sulfate transporter NaS1 (SLC13A1) has the corresponding threonine and proline positions reversed compared with NaDC3, but the P255T/T485P double mutant had no transport activity (Fig. 6). Although the T145A and T527N mutants had low activity compared with the wild type, the signals (26- and 8-fold above background, respectively) were significantly different from the background. Of the remaining mutants, S143A, T253S, P255T, S483A, T485A, T485M, and T485V had between 20–60% of the wild type activity, as expected from their predicted location in the binding site (Fig. 3). Mutations at positions Gly-164 and Cys-517 did not affect succinate transport activity (Fig. 6), consistent with their locations further away from the binding site (Fig. 3).

Substrate Specificity

The substrate specificity of the mutants was tested using two assays: a cis-inhibition assay and the TSR. The cis-inhibition assay measured the transport of [¹⁴C]succinate (10 μm) in the presence or absence of non-radiolabeled test substrates. NaDC3 transport activity was less than 30% of the control in the presence of 1 mm succinate, methylsuccinate, glutarate, α-ketoglutarate, and adipate (Fig. 7), indicating more than a 70% inhibition by these substrates, similar to the results of previous studies (7, 37). The inhibition by citrate was about 20%, verifying a low affinity of NaDC3 for citrate (7). Residues predicted to be in the substrate binding site showed expected changes in substrate specificity upon mutation. For example, T145A, T253S, and T527N had less inhibition by all of the substrates except citrate (succinate, methylsuccinate, glutarate, α-ketoglutarate, and adipate), suggesting a lower affinity for these dicarboxylates (Fig. 7). The P255T mutant only had changes in inhibition by α-ketoglutarate and adipate. Consistent with the different interactions between Thr-485 and substrates in the model (Figs. 3 and 5), mutation of Thr-485 to alanine (found in the rabbit homolog rbNaDC1) or methionine (found in NaCT) resulted in transporters with specific changes in affinity for citrate (Fig. 7). The remaining mutants had similar inhibitory profiles as the WT NaDC3 (data not shown).

As a further test of substrate specificity, we measured the TSR, which provides information on substrate binding during the transition state, an intermediate conformational state during the transport process (34). The TSR is measured using a competitive inhibition assay with two labeled substrates, and it is useful for transporters with low protein expression because it is independent of protein concentration (34). Some of the mutants with changes in substrate specificity in the initial state also exhibited changes in the succinate:glutarate TSR (T145A, T253S, P255T, T485A, and T485M) (Fig. 8B), suggesting alterations in substrate recognition during the transition state. The TSR value for the T527N mutant was difficult to determine accurately because of low activity in the competitive assay, even when measured at 37 °C, although the values appeared to be greater than those of the wild type (Fig. 8). The other mutants had no significant changes in TSR (Fig. 8).

Cation Specificity

As an initial screen of Na⁺ affinity, we examined the transport of succinate at 25 mm and 140 mm Na⁺ to approximate the K_0.5 and saturating concentrations of Na⁺, respectively. Wild-type NaDC3 had ∼50% activity in 25 mm Na⁺ relative to 140 mm Na⁺ (Fig. 9A), consistent with the reported K_0.5 for Na⁺ of 20–45 mm (18, 19). The T145A, T253S, P255T, T485A, T485M, T485V, and T527N mutants all had significantly lower succinate transport activity in 25 mm Na⁺ compared with the WT, suggesting a lower cation affinity. This finding was verified for some of the mutants by measuring the Na⁺ activation of succinate transport (Fig. 9B). The T253S and T485V mutants had Na⁺ activation curves that were shifted to the right, indicating higher K_0.5 values for Na⁺ (Fig. 9B). The S483A mutant had a similar K_0.5 as the WT (17 mm versus 26 mm, respectively) (Fig. 9B). The deleterious effect of the T253S mutation is in agreement with the role of this residue and the N-terminal SNT motif in sodium ion coordination (Fig. 3B). Furthermore, Ser-483 and Thr-485 of the C-terminal SNT motif are located in close proximity to the putative second sodium ion binding site (i.e. Na2), which was suggested by Mancusso et al. (13) for the vcINDY structure (Fig. 10). Thus, the T485V mutation of a polar to a hydrophobic residue likely disrupts key polar contacts. This hypothesis is supported by the MUpro server (PMID 16372356), which predicted the T485V mutation to increase local stability, probably by increasing the hydrophobic effect in this region.

We next examined whether the mutations produced altered cation specificity by measuring transport in the presence of lithium. One of the features of the SLC13/DASS family transporters is the ability to bind lithium at one or more of the three to four cation binding sites (7, 38). In NaDC3, Li⁺ can partly substitute for Na⁺ (Fig. 11 and Ref. 7), producing about 8% of the succinate transport activity in Li⁺ relative to Na⁺. Moreover, the combination of 10 mm Li⁺ and 130 mm Na⁺ resulted in a decrease of succinate transport activity to ∼50% (Fig. 11). Several of the binding site mutants had significantly lower succinate transport activity in the presence of Li⁺ compared with the wild type (S143A, T145A, T485M or V, and T527N). Furthermore, the S143A, T145A, T253S, and T485A,V mutants had less activity than WT NaDC3 in the presence of Li⁺ and Na⁺ together. This result suggests that the binding of lithium does not allow the correct conformational changes to take place to allow optimal succinate binding. Interestingly, the T527N mutant had an increase in activity when the transport buffer contained both Li⁺ and Na⁺, suggesting an alteration in one of the cation binding sites by this mutation (Fig. 11). This result is in agreement with the estimated location of Na2 in close proximity to Thr-527 (Fig. 10).