Structural basis for dynamic mechanism of proton-coupled symport by the peptide transporter POT

To understand the molecular mechanism of H⁺-driven oligopeptide symport by POT, we determined the crystal structures of full-length GkPOT, which consists of 496 amino acids (Fig. 1 A and B). The crystals were obtained using the lipidic cubic phase method. The phases were derived by the single-wavelength anomalous diffraction method, using a selenomethionine-derivatized crystal.

The overall structure of GkPOT includes the N-terminal (H1–H6) and C-terminal (H7–H12) bundles, which are connected by the additional helices HA and HB (Fig. 1 A and B) and share structural similarity with PepT_So and PepT_St (SI Appendix, Fig. S2). A large cleft is formed by the conserved residues between the N- and C-terminal bundles (SI Appendix, Fig. S3). The side chains of these conserved residues are clearly resolved in the electron density map (SI Appendix, Fig. S4A). This cleft contains the peptide binding site opened toward the cytosolic side, thus indicating that the structure represents the inward-open conformation (Fig. 1 A and B). Above the peptide binding site, the extracellular gate is tightly sealed by hydrophobic and hydrogen bond interactions between the N- and C-terminal bundles to prevent H⁺ and substrate leakage (SI Appendix, Fig. S5).

In the crystal structure obtained under conditions containing 300 mM sulfate ion, we observed the electron density of a sulfate ion at the peptide binding site (SI Appendix, Tables S1 and S2 and Fig. S4B). The overall structure of this sulfate-bound form is essentially the same as that of the substrate-free form, with an rmsd of 1.04 Å over all Cα atoms (SI Appendix, Fig. S6). The only structural differences in the peptide-binding site are observed in the side chains of Arg43 on helix H1 and Tyr78 on helix H2, which now hydrogen bond to the sulfate ion (SI Appendix, Fig. S6). The sulfate ion density only appeared when a high concentration (>200 mM) of sulfate was used for crystallization, suggesting that a sulfate ion does not interact with this site under physiological conditions. However, this site coincides with the proposed substrate binding sites in PepT_So (27) and PepT_St (28), raising the possibility that this sulfate ion mimics the C-terminal carboxylate group of a substrate peptide. In this site, the sulfate ion is located near the carboxylate group of Glu310 on helix H7 (SI Appendix, Fig. S4B), which is conserved in the prokaryotic and eukaryotic POT family members (SI Appendix, Fig. S3A). The distance between the oxygen atoms of this sulfate ion and the carboxylate group of Glu310 is 2.7 Å (SI Appendix, Fig. S4B). Considering the low pK_a value for sulfate, the carboxylate group of Glu310 should be protonated to form the observed hydrogen bond. To support this hypothesis, we determined the crystal structures of the E310Q variant (GkPOT-E310Q), which would mimic the protonated state of Glu310, in both the free and sulfate-bound forms (SI Appendix, Tables S1 and S2 and Fig. S7). The structures are almost the same as those of the wild-type protein, and the side chain amide group of Gln310 hydrogen bonds with the sulfate ion (Fig. S7). Therefore, the results support the assignment of the protonation site of the wild-type protein.

Our observations with the sulfate complex led us to attempt to capture a peptide-bound complex with the E310Q variant, using the dipeptide analog alafosfalin, in which the C-terminal carboxylate group is replaced with a phosphonate group (SI Appendix, Fig. S8A). Alafosfalin is a good substrate for POT family transporters (29–31) and competes with the uptake of the (Ala)₂ peptide by GkPOT (Fig. 2A). Therefore, it is likely that naturally occurring peptides also bind to the transporter in a similar manner. The complex structure with alafosfalin was determined at 2.4-Å resolution (SI Appendix, Tables S1 and S2 and Fig. 3 A and B). The overall structure is almost the same as that of the wild-type protein in the sulfate-bound form, with an rmsd of 0.37 Å over all Cα atoms, and the electron density for alafosfalin is clearly resolved within the peptide binding site (SI Appendix, Fig. S8 B and C).

In the complex structure, the phosphonate group of alafosfalin, which corresponds to the C-terminal carboxylate group of the dipeptide, occupies the same position as the sulfate ion in the sulfate-bound structure. The phosphonate group is recognized by the side-chain amide group of Gln310, the guanidinium group of Arg43, and the hydroxyl groups of Tyr40 on helix H1 and Tyr78 on helix H2 (Fig. 3 A and B). The observed configuration suggests that the C-terminal carboxylate group of the dipeptide is also recognized by similar interactions involving the protonated side chain of Glu310. The main chain carbonyl group of alafosfalin is recognized by the hydroxyl group of Tyr40, which also participates in the phosphonate group recognition. The hydroxyl group of Tyr40, in turn, hydrogen bonds with the guanidinium group of Arg36 on helix H1, forming a hydrogen-bonding network (Fig. 3 A and B).

The liposome-based functional analysis revealed that the Ala mutations of Arg36 and Tyr40 abolished both the H⁺-driven uptake and counterflow activities (Fig. 2 B and C), indicating the importance of the interaction network involving Arg36 and Tyr40 for the peptide recognition. A previous functional analysis of PepT_St (28) also showed that the Phe mutation of Tyr30 (Tyr40 in GkPOT) decreased the proton-driven uptake activity, suggesting the importance of the hydrogen bond between the OH group and the substrate. Notably, this interaction network constitutes part of the ExxERFxYY motif, which was previously shown to be essential for peptide transport (19, 28). At the opposite end of the alafosfalin molecule, the N-terminal amino group is recognized by conserved Asn342 on helix H8 and Glu413 on helix H10 (Fig. 3 A and B). The mutation of Glu413 to Gln abolished the H⁺-driven uptake and counterflow activities (Fig. 2 B and C). A conserved Glu resides at an equivalent position on helix H10 in all POT family transporters and is essential for transport in GkPOT, which is consistent with previous observations in PepT1 (32) and the bacterial transporter PepT_St (28). However, the current structure provides a structural insight into the role of this residue in recognizing the N terminus of the bound peptide. In contrast, the mutation of Asn342 to either Ala or Gln had no significant effect on the transport properties of GkPOT (Fig. 2 B and C).

The present high-resolution structure in complex with alafosfalin suggests the recognition mechanism for naturally occurring di- and tripeptide substrates. Previous reports found that the bacterial POT transporters preferentially recognize dipeptides with large and/or hydrophobic side chains (28, 33). In agreement with these previous results, GkPOT recognized (Ala)₂ and (Phe)₂ better than peptides with charged side chains, such as (Glu)₂ and (Lys)₂ (Fig. 2A). The crystal structure revealed the presence of hydrophobic cavities around the side-chain methyl groups of alafosfalin, and these may accommodate the hydrophobic side chains of the substrate peptide (SI Appendix, Fig. S8 D and E). Consistent with previous reports (28, 33), our results showed that GkPOT also recognizes tripeptides, but not peptides longer than 4 amino acid residues (Fig. 2A).

The alafosfalin complex structure revealed additional space on the C-terminal side of the alafosfalin (SI Appendix, Fig. S8 D and E), which could accommodate the third residue in the case of tripeptide binding (SI Appendix, Fig. S9). In addition to the cavities mentioned earlier, small spaces also exist between the substrate main chain and Asn166 on helix H5, Gln309 on helix H7, Asn342, and Glu413 (Fig. 3A and SI Appendix, Fig. S8 D and E). These spaces may collapse in the occluded form of the transporter, which would allow direct hydrogen bonds to form between these protein side chains and the substrate main chain. Supporting this idea, the Ala mutation of Asn166 abolished the H⁺-driven uptake and counterflow activities (Fig. 2 B and C). Therefore, it is possible that the structural transition to the occluded state generates the complete peptide binding site. Further details about the mutant analyses are available in SI Discussion.

To analyze the effects of the H⁺ and peptide binding on the dynamics of GkPOT, we conducted a series of MD simulations in the presence of a 1-Palmitoyl-2-oleoylphosphatidylcholine (POPC) lipid bilayer, based on the present high-resolution crystal structure (SI Appendix, Table S3). First, we examined the effect of the deprotonation of the Glu310 side chain based on a comparison of the 200-ns trajectories of the Glu310-deprotonated and Glu310-protonated simulations (S-E310 and S-E310p, respectively; Fig. 4 A, C, and D). In the S-E310p simulation, no structural change occurred, and the rms fluctuations of both the C- and N-terminal bundles were about 1.0 Å (SI Appendix, Fig. S10). No large rearrangement of the side-chain interactions occurred, supporting our assignment of the protonation states, including that of Glu310. Moreover, these observations suggest that the Glu310-protonated state is stable in the lipid bilayer environment.

In contrast, in the S-E310 simulation, large structural changes were observed. The two-helix hairpin formed by helices H4 and H5 and their connecting loop (the H4–H5 hairpin) in the N-terminal bundle approached helix H8 in the C-terminal bundle, thereby partially occluding the substrate binding site from the cytosolic side (Fig. 4D). This structural motion consists of two components. The first component is the bending of the H4–H5 hairpin. During the simulation, helices H4 and H5 were slightly bent at Pro137 (helix H4) and Pro173 (helix H5) (SI Appendix, Fig. S11A). As a consequence, the H4–H5 hairpin was inclined toward the C-terminal bundle side (SI Appendix, Fig. S11B). The second motion is the domain-level motion between the N- and C-terminal bundles. Compared with the crystal structure, the overall orientation of the N-terminal bundle was rotated by ∼10 degrees toward the C-terminal bundle (SI Appendix, Fig. S12A), which pushed the H4–H5 hairpin further toward the C-terminal bundle and narrowed the cytosolic entrance of the central cleft. Furthermore, these motions brought the N- and C-terminal bundle faces close together in the substrate binding site (SI Appendix, Fig. S13A). During the 200-ns simulation, the occluded form converted back again to the inward-open form (Fig. 4A), indicating that the transporter is in equilibrium between the inward-open and occluded forms in the Glu310-deprotonated state (SI Appendix, Fig. S1 G and H). Another run of the S-E310 simulation also revealed a similar structural change (Fig. S13C), indicating its reproducibility. The S-E310 simulation revealed several interactions that were not observed in the present crystal structures, including the salt bridge between Arg43 and Glu310 (SI Appendix, Fig. S13D). These results suggest that the deprotonation of Glu310 resulted in the salt bridge formation between Arg43 and Gln310, thereby allowing the N- and C-terminal bundles to approach each other.

We then performed the simulation starting from the Glu310-protonated GkPOT in complex with the (Phe)₂ dipeptide (S-E310p-FF). The (Phe)₂ dipeptide was modeled in the substrate binding site according to the results of the simulation with the (Ala)₂ dipeptide (the detailed modeling procedure is provided in SI Appendix, Supplementary Results and Discussion). In the S-E310p-FF simulation, a large structural change similar to that in the S-E310 simulation was detected (Fig. 4 B and E), and the kinks of helices H4 and H5 at the Pro residues and the overall rotation of the N-terminal bundle were both observed (Fig. 3B and SI Appendix, Fig. S12B). The H4–H5 hairpin approached helix H8, and the (Phe)₂ substrate was occluded from the cytosolic side (Fig. 4E). The docked dipeptide was stably bound to the putative substrate-binding site, and no large deviations in the interaction manner were observed. The hydrogen bond between the protonated Glu310 and the carboxylate group of the dipeptide was maintained, and the salt bridge between Glu413 and the amino group of the dipeptide was also stably maintained throughout the simulation (SI Appendix, Fig. S14). The C-terminal phenyl group formed hydrophobic interactions with the hydrophobic pocket (i.e., Tyr78, Trp306, and Trp440), whereas the N-terminal phenyl group was bound to the cavity around the substrate binding pocket (SI Appendix, Fig. S8E). As a consequence, these interactions brought the N- and C-terminal bundle faces close together in the substrate binding site (SI Appendix, Fig. S13B). Therefore, these observations suggest that peptide binding to the Glu310-protonated state changes the electrostatic environment in the central cavity, which may enable the structural conversion between the inward-open and occluded forms (SI Appendix, Fig. S1 E and D). Attenuated total reflection Fourier-transform infrared spectroscopy also detected the structural perturbation of the TM helices, supporting the dynamics predicted from the MD simulations (SI Appendix, Fig. S15 and Supplementary Results and Discussion).

To support the importance of Arg43 and Glu310, we performed the dipeptide uptake and counterflow assays of their mutants. The results showed that the R43Q mutant lost the H⁺-driven uptake activity, but not the counterflow activity (Fig. 2 B and C). These observations indicate that this mutation impaired another state transition than those involving the occluded, substrate/H⁺-bound state (SI Appendix, Fig. S1D). We hypothesize that this mutant impaired the state transition toward the occluded, apo state (SI Appendix, Fig. S1H). The results of the S-E310 MD simulation suggest that the salt bridge is formed between the Arg43 and Glu310 side chains in the transition to the occluded, apo state (SI Appendix, Fig. S13D), but its formation may be impaired in the R43Q mutant. The weaker hydrogen-bonding interaction between Gln43 and Glu310 could not be a driving force to bring the N- and C-terminal bundles close together, resulting in the loss of the H⁺-driven uptake activity.

The pH profile of the R43Q mutant revealed that its H⁺-driven uptake activity slightly recovered under high-pH conditions (SI Appendix, Fig. S16). This observation is also consistent with our hypothesis. In the wild-type protein, the positive charge of Arg43 may facilitate the deprotonation of Glu310 in the transition to the occluded, apo state, which would allow the salt bridge formation between Arg43 and Glu310, as observed in the S-E310 MD simulation (SI Appendix, Fig. S13D). Namely, Arg43 modulates the pK_a of Glu310. This role may be critical, as a protonated form of Glu is more stable than its ionized form in a low-dielectric environment within the membrane. In contrast, in the R43Q mutant, the neutral Gln43 cannot facilitate the deprotonation of Glu310. This may result in the strongly protonated Glu310, thus impairing the transition to the occluded, apo state. Furthermore, the high-pH buffer conditions facilitated the deprotonation of Glu310, which resulted in the slight recovery of the H⁺-dependent uptake activity.

In contrast, the E310Q mutation inactivated the transport in both assays (Fig. 2 B and C). These observations suggest that the H⁺ translocation occurs during a state transition, as in the case of LacY, and that the H⁺ binding site in the outward-open state is different from that in the inward-open state (i.e., Glu310). In this mutant, the transporter in the inward-open conformation can bind the substrate by mimicking the H⁺ bound state of Glu310; however, the H⁺ translocation to the other H⁺ binding site never occurs. Thus, this mutation may block the structural transition between the inward- and outward-open states, resulting in the loss of both the dipeptide uptake and counterflow activities.

Combining the present high-resolution crystal structures, the MD simulations, and the biochemical and biophysical analyses, we can propose a plausible coupling mechanism between the structural transition and the substrate/H⁺ binding in the symport cycle of GkPOT (Fig. 5). In this mechanism, Arg43 and Glu310 play pivotal roles in the substrate/H⁺ coupling: both (i) the protonation of Glu310 and the substrate binding to Arg43 and Glu310 (Fig. 5A) and (ii) the deprotonation of Glu310 and the salt bridge formation between Arg43 and Glu310 (Fig. 5B) induce the structural transition between the inward-open and occluded states. In contrast, the structural transition cannot occur when only H⁺ is bound to the transporter, as the weak interaction between Arg43 and protonated Glu310 could not be the driving force to bring the N- and C-terminal bundles close together (Fig. 5C). In addition, the substrate peptide cannot bind to the peptide-binding pocket in the Glu310-deprotonated form because of the electrostatic repulsion between the carboxylate group of the peptide and the negative charge on the deprotonated Glu310 in the binding site (Fig. 5D). This model clearly explains why the structural transition is allowed when either nothing or both the substrate/H⁺ are bound to the transporter, which is the central basis of the alternating access model for secondary active transporters. It is interesting to note that a similar coupling mechanism was proposed for a fungal high-affinity phosphate transporter (10).

Although our hypothesis can explain the half-cycle of the symport mechanism, the other half of the symport cycle (i.e., the transition mechanism between the outward-facing and occluded forms) remains enigmatic (SI Appendix, Fig. S1). In this cycle, the highly conserved Glu32 on helix H1, which is part of the ExxERFxYY motif, could be involved in the structural transition between the outward-open and occluded states, as its mutation to Gln also abolished both the substrate uptake and counterflow activities (Fig. 2 B and C). One possible explanation is that Glu32 is another protonation site, as discussed earlier, and that the H⁺ translocation occurs between the Glu32 and Glu310 side chains in a manner similar to the H⁺ translocation from His322-Glu269 to Glu325 in LacY. The complete understanding of the coupling mechanism, including the outward-open state, awaits the high-resolution structure of a POT family transporter in the outward-open and/or occluded states.

Previous nonbiased MD simulations of LacY revealed a spontaneous structural change from the inward-open to the partially occluded state (34, 35). In these simulations, both the rigid-body rotation of the N- and C-terminal bundles and the bending motion of the TM segments were reported. The N-terminal bundle exhibited higher internal mobility than the C-terminal bundle (34), and the bending of helices H4 and H5 in the N-terminal bundle brought them toward helices H10 and H11 in the C-terminal bundle to form the partially occluded conformation. The kink in helix H4, found around Pro123, was suggested to confer higher flexibility to allow large conformational changes (34). The C154G mutation, which is considered to stabilize the inward-open conformation (6), is located at this kink in helix H5. Therefore, it was suggested that helix H5 is important for the structural transition (35). Interestingly, the two kinks in helices H4 and H5 are conserved in GkPOT, and the results of the MD simulations (S-E310 and S-E310p-FF) also exhibited structural transitions with similar dynamics (i.e., the domain-level rotation and the bending motions of helices H4 and H5). Therefore, despite the fact that the LacY and POT families are distantly related in the MFS superfamily, the dynamics of the structural transition may be universally conserved among MFS member transporters.

Although similar overall dynamics were observed in GkPOT and LacY, the important sites for the substrate/H⁺ binding of GkPOT, Arg43 and Glu310, are not conserved in other MFS proteins, including LacY, and thus the proposed mechanism by which proton binding promotes the substrate binding and the structural transition of GkPOT is basically different from that of LacY. This discrepancy might arise from differences in the properties of the substrates: peptides have charged moieties and a large dipole moment, whereas sugars have no formal charge and a small dipole moment. Therefore, it is likely that the mechanism coupling the structural transition and the substrate/H⁺ recognition differ between the MFS subfamilies. The overall TM segment topology shared among the MFS members may provide a common framework for the alternating-access mechanism as “hardware,” whereas the mechanism coupling between the structural change and the substrate recognition may be implemented in the individual amino acid sequences conserved in the subfamilies as “software.”