Structural basis for substrate binding and specificity of a sodium–alanine symporter AgcS

Overall Architecture of AgcS.

A M. maripaludis AgcS was expressed in Escherichia coli and purified in various detergents for crystallization. Although successfully crystallized, AgcS alone produced only anisotropically diffracting crystals. The diffraction quality was poor and could not be improved despite major effort. Fab fragments were then generated and cocrystallized with AgcS to improve crystal contacts (SI Appendix, Figs. S1 and S2A). Crystals of AgcS in a complex with Fab fragment diffracted to ∼3.2 Å. The phases obtained by a combination of molecular replacement, using a generic Fab structure, and experimental phasing from Se-Met labeled AgcS were of high quality, allowing us to unambiguously discern two AgcS molecules and two Fab fragments in one asymmetric unit (SI Appendix, Fig. S2B). In total, two structures of AgcS, one with l-alanine bound and the other with d-alanine bound, were determined (SI Appendix, Table S1).

AgcS forms a roughly cylindrical shape, which is ∼40 Å tall and ∼35 Å in diameter comprising 11 transmembrane (TM) helices connected by short cytoplasmic or periplasmic loops (Fig. 1 and SI Appendix, Figs. S3A and S4). Its N terminus is at the periplasm while its C terminus is cytoplasmic. While the N-terminal helix -1 resides outside of the transmembrane bundle and does not seem to participate in translocation of substrates, AgcS appears to maintain pseudo twofold symmetry even though it contains an uneven number of TMs (Fig. 1A and SI Appendix, Fig. S3B). One half contains TMs 1, 2, 5, 6, and 7 while the second half contains TMs 3, 4, 8, 9, and 10. The TMs vary greatly in their length and angle with respect to the plane of the membrane: some helices such as TM2 and TM8 appear almost perpendicular to the membrane plane while TM10 is heavily tilted. TMs 2, 4, 5, 7, 9, and 10 surround the functional core of AgcS, which is made of TMs 1, 3, 6, and 8. Despite the lack of sequence identity (15.1% to LeuT, 14.3% to ApcT, and 14.8% to AdiC), the overall fold of AgcS is similar to that of LeuT (15, 21⇓–23) from the NSS (neurotransmitter:sodium symporter) family. Superposing the crystal structures of AgcS and LeuT gives an overall root mean square deviation (r.m.s.d.) of 4.5 Å. The present structure of AgcS mirrors a common model of LeuT fold Na⁺ and/or H⁺-coupled symporters (24, 25), including ApcT (14) and AdiC (26⇓–28) of the APC family, BetP (29) of BCCT (betaine/carnitine/choline transporter) family, vSGLT (30, 31) of the SSS (solute:sodium symporter) family, and Mhp1 (32) of NCS1 (nucleobase:cation symporter-1) family.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Overall structure of AgcS. (A) The topology of AgcS. The position of the bound alanine is depicted as a yellow triangle. (B) View in the plane of the membrane. Coloring and numbering of helices (cylinders) are the same as in A.

Notably, as with other LeuT-like symporters, two of the functional core helices in AgcS are kinked (Fig. 2, Inset). TMs 1 and 6 line the substrate translocation pathway in AgcS and are broken into two segments. In TM1, Ile-76 and Gly-77 adopt an extended, nonhelical conformation, linking segments 1a and 1b. TM6 contains a longer nonhelical region from Glu-276 to Ser-281 connecting segments 6a and 6b. The residues at these nonhelical loops expose carbonyl and amide groups at the center of the membrane bilayer for hydrogen bonding and ion coordination, which are involved in substrate binding and transport.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

AgcS in the occluded state. AgcS structure was compared with LeuT open-outward state (Left, Protein Data Bank [PDB]: 3tt1, cyan), LeuT occluded state (Center, PDB: 2qei, green), and LeuT open-inward state (Right, PDB: 3tt3, orange). Key structural changes are marked with red arrows. A close-up view of the unwound regions splitting TMs 1 and 6 highlighting residues that are involved in substrate binding is shown in the Inset.

AgcS Structure in an Occluded State.

It is generally accepted that secondary transporters function through an alternate access model (33). The LeuT-fold transporters naturally alternate among three major conformations during their functional cycle: outward-open, occluded, and inward-open state. In the outward-open state, a hydrophilic tunnel emerges and connects the substrate binding site located in the center of transmembrane helical bundle with the periplasm. When in the inward-open state, a hydrophilic tunnel leads from the substrate at the central binding site to the cytosol. During the occluded state, the substrate binding site is isolated from either the periplasm or cytoplasm. Within a complete switch cycle, substrates can be transported across the lipid membrane as the transporter alternates from one conformation to the others.

Two structures of AgcS bound with l-alanine or d-alanine were solved in this study and appeared to be both at the occluded state (Fig. 2). No solvent accessible tunnels or vestibules were apparent leading from the substrate binding pocket across the membrane portion of the transporter. Previous studies reported the structures of LeuT in the open-outward, occluded, and open-inward states (22, 23). APC superfamily transporters such as AgcS and LeuT appear to gate through movements of the functional core helices. As discussed above, these helices are broken (Fig. 2, Inset) to allow free movement of the top half of the helices with respect to the bottom half and this in turn opens the binding pocket either to the outside or the inside of the cell. The pivot point appears to be at the substrate binding pocket. In the overlay of AgcS with the open-outward LeuT (Fig. 2, Left), or the overlay of AgcS with the open-inward LeuT (Fig. 2, Right), core helices deviate by more than 20° or 45°, respectively. No such deviation was observed when the occluded LeuT was overlaid with AgcS (Fig. 2, Center). This analysis clearly suggests that AgcS was captured in the occluded state.

A system of intricate amino acid interactions forms the extracellular and cytoplasmic gates of AgcS. The extracellular gate of AgcS consists of residues from TM1, TM3, and TM8 (Fig. 3A). Specifically, Thr-78 and Ala-73, on the two ends of the unwound region in TM1, expose their Cα nitrogen and carbonyl oxygen atoms for hydrogen bonding with Gln-170 (TM3) and Thr-366 (TM8). On the cytoplasmic side of the binding pocket, the main chain carbonyl oxygen atoms of Ala-74 (TM1) and one of the carboxylate oxygen atoms of Glu-276 (TM6) accept an additional hydrogen bond from the nitrogen atom of indole ring of Trp-370 (TM8), forming a zipper-like motif and adopting the occluded state.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Gating interactions and substrate binding site in AgcS. (A) A close-up view of gating interaction network around the binding pocket in AgcS with l-alanine bound. (B) Hydrogen bonds between amino group of the substrate and the residues of AgcS. (C) Hydrogen bonds and ionic interactions in the hydrophilic part of binding pocket are depicted as dashed lines. Bound Na⁺ ion shown as a gray sphere. (D) The uptake ability for missense mutant (triple alanine mutation at N80A, S274A, and D308A) compared with that of the WT protein.

AgcS Substrate Binding and Ion Modulation.

AgcS binds glycine and both enantiomers of alanine, while strictly excluding other amino acids that were assayed and its transport activity is sodium coupled (SI Appendix, Figs. S5 and S6). In the current study, both structures of AgcS with alanine bound reveal a small amphipathic substrate binding cavity lined by residues from TMs 2, 4, 7, and 8. These residues define two distinct regions in the binding pocket: a hydrophilic charged region close to the extracellular gate and a hydrophobic region embedded deeply in the pocket toward the cytoplasmic gate. A spherical density that could fit the substrate alanine was observed in the composite omit maps at the center of this putative binding pocket (SI Appendix, Figs. S7–S9). Because the resolution of the structures was relatively modest, we relied on a superposition of AgcS with LeuT for placement of the substrate into the binding pocket of AgcS and for confirming the position of the bound Na⁺ ion. A similar approach was used for d-alanine.

The hydrophilic region of the binding pocket interacts with both the amino and carboxyl groups of the bound alanine as well as a bound sodium ion. The amino group of the bound alanine is coordinated by the main-chain carbonyl oxygens of Ala-74, Thr-75 in TM1, and Ser-274 in TM6, and a side-chain carboxyl group from Glu-276 in TM6 (Fig. 3B). Notably, Glu-276 is an ionizable residue whose protonation state may affect substrate binding or transport although this was not investigated in this study. On the opposite side, the exposed main-chain of Gly-77 and Thr-78, which are in the unwound region of TM1, make contacts with the carboxyl group of alanine by direct hydrogen bonds, along with side-chain amide of Gln-170 (TM3), which is also fully conserved (SI Appendix, Fig. S4). This glutamine may also stabilize the irregular structure around the unwound region in TM1 by interaction between its side-chain carbonyl oxygen to the main-chain amino group of Thr-78 (Fig. 3C).

The carboxylic group of the bound alanine is further coordinated to Asn-80 (TM1), Ser-274 (TM6), and Asp-308 (TM7) through the salt-bridge interactions mediated by a single sodium ion (Fig. 3C). These residues are highly conserved in the AgcS family (SI Appendix, Fig. S4). The coordination of the sodium ion is provided by the side-chain carbonyl oxygens of Asn-80 and Asp-308 and the hydroxyl oxygen of Ser-274, further stabilized by main-chain carbonyl oxygen of Ser-274 and Thr-75 (Fig. 3C). The proposed Na⁺ binding site is a conserved ion-binding site of the LeuT fold (21). It provides the correct geometry and chemistry for binding a Na⁺ ion. Uptake studies confirmed that the transport of alanine is Na⁺ coupled (SI Appendix, Fig. S6). To characterize the role of the conserved residues in sodium ion coordination, we generated a triple mutant [triple alanine mutations (3A) at N80A, S274A, and D308A] and evaluated the impact on substrate uptake using proteoliposomes. The mutant markedly reduced the uptake ability of l-alanine, suggesting that the sodium ion plays a key role in facilitating substrate binding and uptake (Fig. 3D).

Substrate Selectivity in AgcS.

The presence of the substrates in AgcS is supported by electron density in the 2mFo-DFc (SI Appendix, Figs. S7–S9) maps as well as the simulated annealing omit maps for both l- and d-alanine (SI Appendix, Fig. S9), but their absolute location and orientation cannot be established due to the limited resolution of the data. The relative positions of the l- and d-substrate also turn out to be difficult to determine: even though the unit cells for the l- and d-alanine structures and the coordinates of the atoms in the corresponding protein molecules are almost identical (SI Appendix, Table S1; all-atom r.m.s.d. 0.18 Å), the AgcS-Fab complexes are at slightly different positions in their respective unit cells (relative translation 0.49 Å). This precludes using an Fo-Fo map to elucidate the differences between the location and orientation of the l- and the d-alanine molecules. There are no significant peaks in the mFo-DFc map for l-alanine, but a pair of opposing positive and negative peaks in the Fo-Fc difference density maps (SI Appendix, Fig. S9F) is revealed after rigid-body refinement of the d-alanine model against the l-alanine data to account for the translation. This indicates a slight shift of the l-alanine toward S274 compared with d-alanine.

AgcS has a high selectivity for small amino acids like glycine and alanine with a dependency on sodium ion. We first evaluated AgcS binding affinities to several amino acids using isothermal titration calorimetry (ITC). While heat exchange was identified for the binding of glycine, and both l- and d-alanine, other amino acids that were tested did not elicit a response (SI Appendix, Fig. S5). Uptake studies using reconstituted proteoliposomes likewise indicated impeccable specificity in AgcS for glycine, l-alanine, and d-alanine in the presence of sodium (Fig. 3D and SI Appendix, Fig. S6).

The crystal structures of AgcS with l-alanine or d-alanine bound indicate that the substrate binding site is small and able to accommodate only small amino acids while maintaining an environment favorable to both alanine enantiomers. In contrast to the polar region of the bound alanine (discussed above), the methyl side chain is surrounded within a hydrophobic pocket mainly formed by the side chain of Ile-165 in TM3, Phe-273 in TM6, Thr-366 and Trp-370 in TM8 (Fig. 4 A and B). That hydrophobic pocket could accommodate both alanine enantiomers equally well. The small size of this hydrophobic pocket, shaped by the van der Waals surface of these binding pocket residues, ensures that only glycine or alanine are accommodated while other amino acids are not. In sharp contrast, the substrate binding pocket in LeuT is much larger (Fig. 4C). Correspondingly, LeuT can accommodate a large variety of amino acids, and indeed the substrate selectivity in LeuT is broader. In AgcS, the aliphatic side chains of any amino acid larger than alanine would clash with Ile-165 and Phe-273. Consistent with this postulate, the mutations I165A and F273A show poor substrate selectivity allowing additional amino acids such as l-valine and l-leucine to be efficiently transported by the mutated AgcS (Fig. 4 D–F). These results indicate that the small size of the substrate binding pocket in AgcS is a key determinant in substrate selectivity, a structural feature that may apply to other APC superfamily members.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Substrate selectivity in AgcS is dictated by the size of the binding pocket. Comparison of the hydrophobic pocket of alanine-bound AgcS and LeuT (Protein Data Bank [PDB]: 2qei). (A–C) Van der Waals surfaces for alanine substrate and interacting residues are shown as colored spheres. The binding pocket of AgcS (dashed ring in C) is very small and only glycine or alanine can fit while in LeuT, where substrate selectivity is broader, the site is much larger to accommodate larger amino acids. (D–F) AgcS mutants I165A and F273A lose substrate selectivity allowing valine and leucine to be transported. Nonspecific uptake was assessed by using protein-free liposomes under identical conditions as described in Methods Summary. Ctrl, control.

Most amino acid transporters of the APC superfamily have effective stereo-selectivity and tend to transport l-type amino acids exclusively, mirroring the fact that l-amino acids are most commonly used in life while d-amino acids are rarely used. However, AgcS has an ability to transport both enantiomers of alanine as an important nitrogen source (5, 6). This alanine utilization plays a crucial role to support M. maripaludis in harsh environments. As discussed above, the substrate binding site in AgcS has a cytoplasmic-facing hydrophobic surface used to accommodate the aliphatic part of either l- or d-alanine equally well (Fig. 5 A and B). To test this hypothesis, we generated a mutant, W370Q, and assayed the selectivity of the transporter for l- and d-alanine (SI Appendix, Fig. S6). While the ability of W370Q AgcS to uptake l-alanine did not change much, its ability to transport d-alanine was reduced (SI Appendix, Fig. S6E versus SI Appendix, Fig. S6F). This analysis demonstrated that the polarity of the residues at the substrate binding site affects the ability to differentiate between l- and d-amino acids. In sharp contrast, transporters that only allow l-type amino acids to be transported, like LeuT, do not have an even distribution of hydrophobic surface at their substrate binding site. Instead, the cytoplasmic face of their substrate binding site is amphipathic containing both a hydrophobic patch and a charged polarized patch which is also space-limited (Fig. 5C). This feature ensures that only the l-enantiomers of amino acids can bind while d-enantiomers are repelled electrostatically.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Stereo-selectivity is afforded by electrostatic repulsion. The amide plane of substrate is shown as gray dash. (A and B) In AgcS both alanine enantiomers fit equally well as the two faces of the binding pocket are hydrophobic. (C) In LeuT, which only binds l-amino acids, one face is hydrophobic while the other is polarized ensuring selection for l-amino acids over d-amino acids.

Concluding Remarks.

The mechanism of substrate binding and selectivity is an essential feature for the APC superfamily members. Here we described the crystal structure of an alanine transporter AgcS capable of transporting both l- and d-alanine and discovered that both the size and polarity of the substrate binding pocket are important structural determinants for specificity and stereo-selectivity. While the resolution of these structures was modest, superposition with LeuT allowed us to identify the position of the bound alanine and that of the Na⁺ ion. The size of the binding pocket in AgcS determines what amino acids could fit in the transporter while electrostatics control whether l-type or d-type amino acids could bind. We showed that a single point mutation in the binding site of AgcS can broaden the specificity of this transporter to allow uptake of larger amino acids such as valine and leucine. Further, introduction of a charged residue instead of the tryptophan at the binding pocket interfered with the transporters ability to transport d-alanine.