Protonation of Glu¹³⁵ Facilitates the Outward-to-Inward Structural Transition of Fucose Transporter

Abstract

Major facilitator superfamily (MFS) transporters typically need to alternatingly sample the outward-facing and inward-facing conformations, in order to transport the substrate across membrane. To understand the mechanism, in this work, we focused on one MFS member, the L-fucose/H⁺ symporter (FucP), whose crystal structure exhibits an outward-open conformation. Previous experiments imply several residues critical to the substrate/proton binding and structural transition of FucP, among which Glu¹³⁵, located in the periplasm-accessible vestibule, is supposed as being involved in both proton translocation and conformational change of the protein. Here, the structural transition of FucP in presence of substrate was investigated using molecular-dynamics simulations. By combining the equilibrium and accelerated simulations as well as thermodynamic calculations, not only was the large-scale conformational change from the outward-facing to inward-facing state directly observed, but also the free energy change during the structural transition was calculated. The simulations confirm the critical role of Glu¹³⁵, whose protonation facilitates the outward-to-inward structural transition both by energetically favoring the inward-facing conformation in thermodynamics and by reducing the free energy barrier along the reaction pathway in kinetics. Our results may help the mechanistic studies of both FucP and other MFS transporters.

Introduction

As one of the largest membrane protein families, major facilitator superfamily (MFS) proteins transport a large variety of substrates including many fundamental life elements (e.g., ions, sugars, and bases), and thus play indispensable roles in sustaining life (1–3). To date, >10,000 proteins have been identified in the MFS family, most of which are supposed to adopt the same structural topology despite limited sequence identities (4). A typical MFS member is composed of 12 transmembrane helices, among which the first six helices (TM1–6) and the last six ones (TM7–12) constitute two separate domains (5). Occasionally, two or more additional helices may appear between the two domains (6–10). Based on the crystallographic studies (6,11–16), MFS proteins show an inverted-repeat feature, and are thus supposed to be evolved by gene duplication at an early stage (17). Structure and sequence alignments indicate that MFS transporters may evolve by a mix-and-match mechanism, and that functional sites from different transporters can coincide with each other (18). Therefore, MFS transporters are believed to share a universal transport mechanism (17,19), and study on one member is of great reference to the others.

As potential pharmaceutical targets, MFS transporters have been extensively explored in both structures and functions (20–22). The first high-resolution structure came from the lactose permease in Escherichia coli (LacY), which was originally known as a member in the lac operon in 1960s (23) but quickly became a hotspot for mechanistic study after its structural determination. The crystal structure of LacY exhibits a substrate-binding pocket accessible to the cytosol and this conformation is therefore named as inward-open (13,24). According to the alternating access model (17), transporters require both outward-open and inward-open conformations as well as some intermediate states (e.g., the occluded conformations) to fulfill a complete cycle of substrate transport. Until now, crystal structures from several orthologs have been successively determined, and the full transport cycle of MFS transporters could be presumably assembled from these homologous structures (6,8,9,12–16,24–26).

The L-fucose/H⁺ symporter (FucP) is the first MFS member whose crystal structure is determined in an outward-open conformation (25). Following the typical MFS topology, the 12 transmembrane helices (TM1–12) in FucP can be assigned to an N-terminal domain (NTD) and a C-terminal domain (CTD), which are supposed to move as rigid bodies during the substrate translocation based on structural comparison between FucP and LacY (25). In the periplasm-accessible vestibule sandwiched by the two domains, a β-NG (n-nonyl-β-D-glucopyranoside) presents a sugar headgroup to closely interact with Glu¹³⁵ and Asn¹⁶², both of which are indispensable residues based on mutagenesis studies (25). The binding pocket that holds the headgroup of β-NG is therefore believed to be the native binding site for fucose (25). Numerous residues have been proposed as essential for substrate transport based on the structure and extensive biochemical mutagenesis analysis (18,27,28). Among these residues, the ionizable residue Glu¹³⁵ located in the central vestibule is the most important one, not only because it is directly responsible for fucose binding, but also because it is supposedly involved in H⁺ translocation and protein conformational change (25).

Although the crystal structures of the numerous MFS transporters available as of this writing have already exhibited nearly all conformational states required for substrate transport, the observation of the structural transition for one particular protein target is still lacking. In this work, we studied the conformational change of FucP using molecular-dynamics simulations. We not only observed the large-scale outward-to-inward structural transition, but also evaluated the free energy change during this process. Our results illustrated the critical contribution of Glu¹³⁵ in the coupling between H⁺ translocation and conformational change: the protonation of Glu¹³⁵ facilitates the outward-to-inward structural transition both by energetically stabilizing the inward-facing conformation and by reducing the free energy barrier along the reaction pathway.

Materials and Methods

Numerous simulation techniques were combined in this work to comprehensively investigate the molecular mechanism of FucP. At the beginning, two parallel equilibrium simulations were performed to evaluate the structural fluctuation of FucP with Glu¹³⁵ protonated and deprotonated, respectively. Subsequently, the adaptive tempering (AT) simulation (29) was engaged to enhance the sampling over the complex conformational space for both systems. Based on the AT trajectories, a series of simulations were then conducted following the protocol in Fig. S1 in the Supporting Material to estimate the free energy landscapes and identify the reaction pathways. In specific, the two-dimensional plane composed of a pair of order parameters was first coarsely gridded and one low-energy-representative snapshot for each bin was selected from the AT trajectory (Fig. S1 A). Two rounds of equilibrations at 310 K were then conducted over these representative snapshots, which can both effectively rescue the structural distortion caused by heating during the AT simulation and raise the sampling density in all accessible regions on the map (Fig. S1, B and C). Finally, the two-dimensional plane was finely gridded and the conformation closest to the center of each bin was chosen as the initial structure for the evaluation of the potential of mean force (PMF), which eventually generated the two-dimensional PMF profile to allow the identification of reaction pathway (Fig. S1 D).

Basic information for simulation and postsimulation analysis

All simulations were prepared with the program VMD (Ver. 1.9.1; http://www.ks.uiuc.edu/Research/vmd/) (30), and were conducted using the program NAMD (Ver. 2.9; http://www.ks.uiuc.edu/Research/namd/) (31). The CHARMM27 force field (http://www.charmm.org/) (32–34) was adopted, if not otherwise stated. The van der Waals interactions were cut off at 12 Å, with a smooth switch at 10 Å. The electrostatics was estimated using the particle-mesh Ewald method with the periodic boundary condition applied in all directions (35–37). The principal component analysis (PCA) was performed using the NMWIZ tool (http://www.findbestopensource.com/product/nmwiz) within the protein dynamics analysis toolbox PRODY (http://prody.csb.pitt.edu/manual/getprody.html) (38). The pore radii for membrane proteins along the axis perpendicular to membrane surface were calculated using the software HOLE (Ver. 2) (39,40). Hydrogen bonds were identified using VMD, with the distance and angular cutoffs set as 3.5 Å and 40°, respectively. The pK_a values were estimated using the downloadable program PROPKA (Ver. 3.1; http://159.149.85.2/propka_about.htm) (41,42) with all parameters set to the default values. All simulations conducted in this work are listed in Table S1 in the Supporting Material and the total simulation time exceeds 7 μs.

Equilibrium simulations

Two conventional equilibrium simulations were conducted to analyze the structural fluctuation of FucP at different protonation states of Glu¹³⁵. The initial outward-open structure was downloaded from the Protein Data Bank database (PDB; http://www.wwpdb.org/) (43) (PDB:3O7Q (25)). The ionizable residue Glu¹³⁵, which is located in the vestibule sandwiched between NTD and CTD, is critical for the cotransport of substrate and proton, as suggested by both in vivo and in vitro mutagenesis assays, where its mutation to Ala or Gln completely blocks substrate transport (25). Although this residue is suspected to be responsible for H⁺ translocation, no conclusive evidence, however, can verify this hypothesis so far. Therefore, two systems were designed to reflect the protonation state of Glu¹³⁵, denoted as 135P and 135NP for protonated and deprotonated Glu¹³⁵, respectively. The suffix of “sub” was added to represent the presence of bound substrate in the initial structure. Because Glu⁴¹⁵ is buried in a hydrophobic environment, and because calculation using the software PROPKA also confirms its high pK_a value of 9.71, this residue was set to the protonated state to stabilize the protein structure. The protonation states of all the other ionizable residues were set to the default options.

The protein was inserted into the POPE bilayer (containing ∼217 POPE molecules), and the substrate L-fucose was added to the systems based on the position of the β-NG headgroup in the crystal structure. After solvated by ∼18,786 TIP3P waters, 0.1 mol/L NaCl were added in the systems to mimic the physiological condition. The numbers of Na⁺ and Cl⁻ ions were precisely controlled to neutralize the systems. Both systems were simulated at 310 K with a time step of 1 fs. Using the same protocol as our previous works (44), 1.5 nanoseconds of step-by-step preequilibrations were conducted to relax the energy in an NVT ensemble, after which the simulations shifted to an NPT ensemble. The pressure was held at 1 atm in the NPT simulations. The first 10 nanoseconds in the NPT trajectories were also regarded as preequilibration and the following frames were collected as the productive simulations.

Enhanced sampling by accelerated simulations

Conventional equilibrium molecular-dynamics simulations can seldom capture the large-scale movement of biomolecules within a manageable time, especially for the conformational change of transporters, which is more time-consuming due to the geometric constraining by membrane bilayers. We therefore adopted the AT method (29) to enhance the conformational sampling. In the AT simulation, the system temperature is automatically adjusted according to the system potential at the previous stages, therefore raising the probability to overcome the energy barrier (29).

Both systems (135Psub and 135NPsub) were simulated to investigate the effects of Glu¹³⁵ protonation on the conformational change, starting from the last conformations derived from the corresponding equilibrium simulations. The minimal temperature was set to 310 K. According to our tests, 500 K is the lowest possible value for the maximal temperature at which significant protein conformational changes can be observed within 30 ns. Therefore, the maximal temperature was set to 500 K to avoid structural distortion caused by overheating as much as possible. The default value of 0.1 was adopted for the tunable variable gamma. The temperature was reassigned every 100 steps, and the simulation time step was 1 fs. If the system temperature reached the upper or lower boundaries (the minimal and maximal temperatures), the temperature reassignment would be disabled for the next few cycles to allow the system temperature to gradually return to the preset range. For each system, 30 nanoseconds of AT simulation were performed in total.

The structural snapshots were described by two order parameters, d_EC and d_IC, which quantify the interdomain distances at the extracellular (periplasmic) and intracellular (cytoplasmic) sides, respectively. In specific, each parameter is defined as the distance between the center of the C_α atoms of six residues (one residue from each of TM1–6) at the corresponding membrane-solution interface of NTD and that for six residues (one residue from each of TM7–12) of CTD on the same interface (see Fig. 1 D). In this way, each protein conformation could be represented as a point on the two-dimensional map of d_IC and d_EC.

Figure 1.

Structural fluctuation in the equilibrium simulations. (A) Time series of C_αRMSD for the 135Psub (black) and 135NPsub (red) systems. (B) Time series of d_EC for the 135Psub (black) and 135NPsub (red) systems. (C) Time series of d_IC for the 135Psub (black) and 135NPsub (red) systems. (D) Schematic plot showing the definitions of d_IC and d_EC. C_α atoms belonging to the NTD (blue spheres) and CTD (yellow spheres) are used in the definition. To see this figure in color, go online.

Equilibrating the snapshots from AT simulations

Conformations in the AT simulations are obtained at various temperatures, and therefore may have severe local structural distortion. To correctly describe the outward-to-inward structural transition and to collect good conformations for the subsequent free energy calculations, trajectories of the AT simulations should be sufficiently equilibrated at the physiological temperature (310 K). The 135Psub system exhibits the complete conformational change in the AT simulation (see the Results for details). Therefore, the selected conformations for this system were relaxed in a standard two-stage protocol (as shown in the next two paragraphs). In contrast, the 135NPsub system is incapable of sampling the inward-facing conformation in the AT simulation. To evaluate the complete structural transition for this system, the initial conformations of this system were mutated from the corresponding protonated conformations of the 135Psub system after the first stage of equilibration. After 10 nanoseconds of preequilibration, these conformations were then further relaxed in the second stage.

In the first stage, all structural snapshots from 135Psub AT trajectories were scattered onto the d_IC-d_EC map (Fig. S1 A), which was then gridded using a lattice of 0.5 Å for both d_IC and d_EC. Among all 0.5 × 0.5 Å² bins, only the 44 high-occupancy ones (containing at least eight structures) were chosen for the next step. The lowest energy structure in each bin was selected to initiate a 10-ns equilibrium simulation. The trajectories were saved every 20 ps, and the first 4-ns preequilibration data were disregarded. Therefore, 13,200 frames were collected in total, and scattered onto the d_IC-d_EC map again (Fig. S1 B).

In the second stage, plenty of relaxed conformations on the d_IC-d_EC map guarantee sufficient sampling along the structural transition pathway. Consequently, the number of high-occupancy bins (containing at least eight frames) rose to 68. For each bin in the 135Psub system, the structure closest to the bin center was selected for another 10-ns equilibrium simulation. As for the 135NPsub system, the starting conformation for each bin was derived from that of 135Psub collected from the first stage, and all structures were further equilibrated for 12 ns after 10 ns of preequilibration. For both systems, the frames of the last 8 ns were kept for further analysis, which produces 26,800 frames in total (Fig. S1 C).

Free energy calculations

After two rounds of equilibration, we had plenty of relaxed conformations along the pathway for both systems. These structures were then fed into simulations performed by the adaptive-biasing-force algorithm (ABF) (45) to estimate the two-dimensional PMF profiles.

All conformations collected after the two-stage relaxation were scattered onto the d_IC-d_EC map with a lattice size of 0.5 Å. The conformation closest to the center of each bin was extracted as the starting structure in the ABF calculation of that bin (Fig. S1 C). The width for ABF presampling was set to 0.02 Å for both d_IC and d_EC, leading to a total of 625 grids in each bin. For the bins where the ABF calculation did not converge after 20 ns of simulation, four additional overlapped bins with shifted center of 0.25 Å were then applied in the directions of both d_IC and d_EC, which thus fully covered the original bin for better convergence. In total, 241 and 293 bins were created for the 135Psub and 135NPsub systems, respectively. ABF simulations were carried out in stages, with the first stage spanning 4 ns, and each of the following ones, 2 ns. Simulation in each bin was conducted for at least three stages (simulation time ≥ 8 ns), and stopped only when convergence was reached.

Convergence was evaluated using the ABF parameter RCOUNT, which was firstly introduced by the ABF developers as the ratio of the maximal and minimal counts of samples in each bin (45). In this study, convergence was believed to be achieved when RCOUNT was less than sixfold in the last eight nanoseconds of simulations. In this way, the subspace in each bin can be sampled nearly uniformly, which allows accurate estimation on the PMF. The overall PMF profiles were spliced from those in individual bins (Fig. S1 D), using the ABF_INTEGRATE program in NAMD (http://www.ks.uiuc.edu/Research/namd/).

Finding the reaction pathway

The detailed pathway for structural transition can be inferred from the calculated two-dimensional PMF profiles (Fig. S1 D). In this work, we borrowed the idea from the program TRACE_IRC (http://www.acmm.nl/ensing/software/index.html) (46), which conducts cyclic searches recursively. In our case, we firstly identified the outward-open (triangle, start conformation), outward-facing partially opened (circle), occluded (diamond), and inward-facing (square, end conformation) conformations by searching the energy minima within preestimated regions (Fig. S1 D). Next, the coarse cyclic search was performed with a searching radius of 1.0 and 0.8 Å for the 135Psub and 135NPsub systems, respectively, which resulted in a coarse pathway from the starting to the ending conformation. The finer pathway was identified by a series of iterative cyclic searches between two successive coarse points with a smaller radius of 0.1 Å. Several combinations of searching radii were tested, and the ones chosen above were shown to better reflect the reaction pathway after visual inspection.

Four tricks should be noted when guessing the pathway. First, the angle between the moving direction and the start-to-end direction should be limited to avoid overtwisting. In our case, the angle should be smaller than 60° and 45° for the coarse and finer searches, respectively. The choice of the angles as well as searching radii is case by case, which requires manual inspection. Second, if the lowest energy point in a new search is coincident with the previous point, the next point is chosen as the collinear point in opposite direction with the given radius. Third, if two successive pathway segments intersect, the points in-between are removed. Fourth, all points falling within a given threshold distance (0.08 Å) to any other existing points will be deleted. (The PYTHON code for pathway identification is downloadable from the website http://166.111.152.91/path_finder.html.)

Results

No significant conformational changes observed in the equilibrium simulations

Theoretically, a membrane transporter has to undergo considerable conformational change to accomplish substrate transport. As a first trial, two conventional equilibrium simulations were carried out to investigate the structural fluctuation of FucP when Glu¹³⁵ is protonated or deprotonated in the presence of the substrate (called 135Psub and 135NPsub, respectively; see Table S1).

As shown in Fig. 1 A, both systems are relatively stable without significant conformational changes. In specific, the C_α root-mean-square distances (C_αRMSDs) for both systems reach the average level of ∼2.5 Å within the first 10 nanoseconds of simulations, and remain steady for the next 90 ns, in spite of slight fluctuations. There is no significant difference between the Glu¹³⁵ protonated and deprotonated proteins.

Here we used two order parameters, d_IC and d_EC, to quantify the degree of conformational change from the outward-facing to the inward-facing states (Fig. 1 D). These variables describe the distances between NTD and CTD at the intracellular (cytoplasmic) and extracellular (periplasmic) membrane-solution interfaces, respectively, and can effectively discriminate various conformational states of the crystal structures determined for MFS transporters (Fig. S2). Similar to the C_αRMSD analysis, both d_IC and d_EC are quite steady, indicating the lack of obvious structural change (Fig. 1, B and C).

In summary, the above results negate the presence of observable functional movement in the equilibrium simulations. According to previous experiments on glutamate and aspartate transporters, the large-scale conformational changes of membrane transporters frequently fall in the timescale of seconds (47,48). Such slow molecular events therefore may escape the observing capability of all-atom equilibrium simulations that can only track the timescale of microseconds or perhaps milliseconds at most.

The outward-to-inward structural transition observed in accelerated simulations

Because conventional equilibrium simulations are incapable of sampling the large-scale conformational change as indicated previously, we adopted an accelerated algorithm, the AT simulation method (29), to enhance sampling in the conformational space (Table S1).

As expected, the AT method raises the magnitude of movement significantly and both systems deviate from their initial conformations substantially (Movies S1 and S2). The Glu¹³⁵ protonated protein (135Psub) has a large conformational change (average C_αRMSD = 4.94 Å) during the 30-ns AT simulation. In particular, the average C_αRMSD for the last 2 ns reaches 7.28 Å (Fig. 2 A), greatly surpassing the equilibrium simulation (2.5 Å, Fig. 1 A). Moreover, the conformation indeed changes toward the inward-facing state (Movie S1), as suggested by the rise of d_IC from 21 Å to nearly 30 Å and the drop of d_EC from 32 Å to ∼21 Å (Fig. 2 B). We clustered all structural frames in the 135Psub trajectory by their C_αRMSDs, and found out that the fifth cluster corresponds to the inward-facing conformation. A representative structure (the one closest to the average conformation) for the cluster was then extracted for further analysis after a short equilibration at 310 K (Fig. S3). Notably, in contrast to the substantial overall C_αRMSD (7.12 Å) from the crystal structure, both the NTDs and CTDs of this structure experience milder changes, as suggested by the C_αRMSDs (for TMs only) of 2.92 and 2.78 Å, respectively. The large overall conformational change is mainly caused by the domain reorientation (rotation by ∼40°, Fig. S3 A) rather than the intradomain movement. The angle of domain rotation was estimated from the quaternion to structurally align the CTDs when the NTDs were superposed first, and the value agrees well with the structural comparison on FucP and LacY (25). In addition, the pore radii for this representative structure were calculated along the axis perpendicular to the membrane using the program HOLE (40). As expected, the periplasmic-half becomes sealed as compared to the crystal structure, while the cytoplasmic-half is expanded to a level to be accessible for the cytosolic water (Fig. S3 B). Unfortunately, the cytoplasmic entrance of this representative structure is not wide enough for the substrate release. Therefore, we described this inward-facing conformation as “partially opened” (referred to later in this article as “inward-facing”). The Glu¹³⁵ deprotonated system (135NPsub) exhibits a similar tendency, but with a smaller scale in terms of the changes of C_αRMSD as well as d_IC and d_EC (Fig. 2). In specific, the conformational change stops at an occluded state (Fig. S3 B), and does not proceed further to reach the inward-facing state (Movie S2).

Figure 2.

Structural transition in the AT simulations. (A) Time series of C_αRMSD for the 135Psub (black) and 135NPsub (red) systems. (B) Time series of d_IC (black) and d_EC (red) for the 135Psub system. (C) Time series of d_IC (black) and d_EC (red) for the 135NPsub system. To see this figure in color, go online.

The distinct behaviors between the Glu¹³⁵ protonated and deprotonated proteins can be easily seen from the scatter plots of structure snapshots on the (d_IC, d_EC) two-dimensional plane, where the outward-facing and inward-facing conformations occupy the upper-left and lower-right corners, respectively (Fig. 3). In the presence of substrate, the Glu¹³⁵ protonated protein (135Psub) can sample the overall plane, while the Glu¹³⁵ deprotonated protein (135NPsub) is less likely to visit the lower-right corner.

Figure 3.

Distributions of structural snapshots on the d_IC-d_EC map for the AT trajectories of the 135Psub (left panel) and 135NPsub (right panel) systems.

It has been reported that the uptake of substrate L-fucose or D-arabinose leads to fluorescence quenching of Trp³⁸ and Trp²⁷⁸ in FucP (27). The quenching is supposedly caused either by the interaction between the pyranose ring of the substrate and the indole ring of Trp, or by the approximation of TM1 and TM7. Both phenomena are observed in the 135Psub trajectory. On one hand, the substrate gradually approaches Trp²⁷⁸ to a short distance (Fig. S4, A and C). In specific, hydrogens on the pyranose ring could approximate the Trp²⁷⁸ side-chain carbons to a distance of ∼2.90 Å, a value that is smaller than the quenching distance of 3.2 Å (49). On the other hand, the periplasmic distance between TM1 and TM7 (quantified by the distance between the C_α atoms of Pro⁵⁰ and Arg²⁸³) decreases from 23.31 to 7.65 Å (Fig. S4, B and C).

In addition, accompanying the protein conformational change, the fucose leaves the native binding site and moves to another location, where it can form hydrogen bonds with Asn⁴³ and Asn¹⁶² and bury its hydrophobic methyl group within the side chains of three aromatic residues, Tyr³⁶⁵, Phe³⁰⁸, and Trp²⁷⁸ (Fig. S5). To identify these residues, all frames from the AT trajectory of the 135Psub system were clustered based on the position of fucose after aligning the protein structures, and the representative structure (centroid frame) of the largest cluster was extracted first for visual inspection. Subsequently, the identified interactions were validated by counting their percentages within the cluster. Based on these analyses, the fucose forms hydrogen bonds with Asn⁴³ and Asn¹⁶² in 64.09 and 40.54% of all cluster frames, respectively. On the other hand, the hydrophobic methyl group of fucose is buried among the side chains of Tyr³⁶⁵, Phe³⁰⁸, and Trp²⁷⁸ (distance between group centers ≤ 6.0 Å) in 41.3, 44.8, and 17.8% of all cluster frames, respectively. Coincidently, experimental data show that Asn¹⁶², Tyr³⁶⁵, Phe³⁰⁸, and Trp²⁷⁸ are all critical for the transport of substrates (18,25).

The free energy landscape and reaction pathway identified for the structural transition

To better understand the thermodynamics and kinetics during the outward-to-inward structural transition, we performed ABF calculations (45) to estimate the free energy landscapes in the space of the order parameter pair (d_IC, d_EC) (Table S1), after systematic structural equilibrations on the structural snapshots from the AT simulations (Fig. S1). Notably, the perturbation on membrane lipids introduced by the heating in the AT simulation can be effectively removed after equilibration at 310 K (Fig. S6), which enables reliable estimation of the free energy for membrane transporters.

In general, the 135Psub and 135NPsub systems exhibit similar shapes in the two-dimensional PMF profiles (Fig. 4), with the local minima around coordinate (21.5, 30.5; State 1) corresponding to the crystal structure. The representative conformation of this state indeed exhibits an outward-open conformation (Figs. S7 A and S8). Both PMF profiles also show local minima at the lower-right corner around the coordinate (26.5, 20.0; State 4), which correspond to the inward-facing state. Although the representative structure only possesses a small cytosolic opening, its central vestibule is indeed accessible to cytosolic waters (Figs. S7 D and S8). Besides these two states, two other local minima identified around the coordinates (22.0, 26.0; State 2, and 23.5, 23.0; State 3) may reflect the intermediate states during the structural transition. The latter intermediate state adopts the occluded conformation while the former one takes the outward-facing partially opened structure (Figs. S7, B and C, and S8). Indeed, the free energy landscapes do not contain the complete inward-open state, possibly due to the insufficient sampling of the AT simulations. More enhanced sampling is possibly required to capture such a state, which may alter the final free energy landscape by introducing new free energy minima and/or barriers.

Figure 4.

Free energy landscape and the structural transition pathway. (A) The two-dimensional PMF profiles for the 135Psub system, with reaction pathway (red line). (B) The two-dimensional PMF profiles for the 135NPsub system, with reaction pathway (red line). (C) Overlay of the pathways of the 135Psub (black) and 135NPsub (red) systems on the same plot. (D) The one-dimensional along-the-pathway PMF profiles for the 135Psub (black) and 135NPsub (red) systems. (Triangle, circle, diamond, and square) Free energy minima according to States 1–4, respectively. The scales of free energy values for the two-dimensional PMF profiles are shown in the color bar, in the units of kcal/mol. To see this figure in color, go online.

The reaction pathways for the structural transitions can be identified on the two-dimensional PMF profiles. As shown in Fig. 4, A and B, the pathway estimation is quite reasonable, because the identified paths travel through most energy minima and pass the high-energy regions through the saddle points (e.g., around the coordinate (26.5, 21.5) in the 135Psub system). As shown in Movie S3, the conformational change along the identified pathway is much smoother than those changes observed in the AT simulations (compare with Movies S1 and S2). In a further validation test, we collected all conformations from the ABF simulations and performed a PCA analysis to check the continuity of sampling along the reaction pathway (Fig. S9). Only the first two principal-component modes were finally retained to reflect the global conformational change (with contribution >10%): PC1 exhibits high correlation with both d_IC and d_EC (with correlation coefficients = 0.82 and −0.93, respectively) and therefore clearly describes the outward-to-inward structural transition; PC2 represents the major global movement that is orthogonal to the reaction pathway. A scatter-plot of the ABF snapshots in the two-dimensional plane of PC1 and PC2 (see Fig. S9) indicates that the major orthogonal degree of freedom (PC2) is continuously sampled in the ABF simulation, which thereby supports the proper free energy calculation and reaction pathway identification. As shown in Fig. 4, although the final reaction pathways identified for the 135Psub and 135NPsub systems exhibit similar shapes (Fig. 4 C), the one-dimensional, along-the-pathway PMF profiles, are quite different (Fig. 4 D). Estimating the statistical errors for the free energies from the ABF simulations is very expensive. Therefore, we only provided a rough estimation, based on the three two-dimensional PMF profiles derived at the final time step and at the time steps of 2 and 4 ns before. The one-dimensional PMF profiles with error bars are shown in Fig. S10.

Protonation of Glu¹³⁵ favors the outward-to-inward structural transition

According to previous experimental data, the mutation of Glu¹³⁵ nearly completely destroys the substrate transport, in both the in vitro counterflow assay and the in vivo test (25). Although Glu¹³⁵ has been confirmed as essential in substrate binding, its role in the H⁺ translocation and protein conformational change is speculative, and awaits experimental validation.

Here, to further investigate the relationship between Glu¹³⁵ and H⁺ translocation as well as protein conformational change, we compare the one-dimensional PMF profiles along the reaction pathway for the 135Psub and 135NPsub systems. As shown in Fig. 4 D and Fig. S10, the curves of both systems start from the outward-open state (triangle), then sequentially visit the outward-facing partially opened state (circle) and the occluded state (diamond), and finally reach the inward-facing state (square). In the 135Psub system, the starting and ending states show roughly identical free energy values, with the former slightly preferred by 0.51 kcal/mol, a value smaller than thermal fluctuation of kT, where k is the Boltzmann constant and T is the absolute temperature. In contrast, the free energy difference between the two states rises considerably to 2.70 kcal/mol in the 135NPsub system. The protonation of Glu¹³⁵ thus specifically favors the inward-facing conformation by 2.19 kcal/mol, relative to the outward-open state.

Moreover, the shapes of the along-the-pathway PMF profiles show different kinetics during the structural transition for the two systems. In the presence of substrate, the protein with Glu¹³⁵ deprotonated (135NPsub) naturally transits to an outward-facing partially opened conformation in a downhill manner. From this state, the protein has to overcome a small free energy barrier of 2.52 kcal/mol to reach the occluded conformation, which is slightly less stable by 0.90 kcal/mol. Further structural transition from the occluded state to the inward-facing one is precluded by a huge free energy barrier of 6.56 kcal/mol. For this system, the overall free energy barrier from the most stable intermediate to the final state is 7.46 kcal/mol, in total. In contrast, for the Glu¹³⁵ protonated protein (135Psub), the most stable state along the pathway is the occluded conformation. In the presence of substrate, the protein can easily reach this state by overcoming tiny barriers of ∼1 kcal/mol. Although a free energy barrier of 5.76 kcal/mol hinders the ongoing structural transition to the inward-facing conformation, the overall barrier along the pathway is smaller, by 1.70 kcal/mol, than the value for the Glu¹³⁵ deprotonated system.

In summary, the protonation of Glu¹³⁵ facilitates the outward-to-inward transition both by biasing the thermodynamic equilibrium toward the inward-facing conformation and by lowering the overall free energy barrier to accelerate the structural transition in kinetics. These observations agree with the AT simulations well (see Fig. 3).

Discussion

Protonated Glu¹³⁵ stabilizes the inward-facing conformation

As stated previously, the protonation of Glu¹³⁵ stabilizes the inward-facing conformation by ∼2 kcal/mol, relative to the outward-open state. To investigate the reason, we took all structural snapshots within 1 Å around the free energy minimum of the inward-facing state (State 4 on the d_IC-d_EC map) from the second-round equilibration trajectories of both systems and performed structural analysis on the largest structural cluster (generated by pairwise C_αRMSD) of these conformations.

We first compared the representative structures (the frames closest to the average structures within the cluster) of the two systems. Surprisingly, they do not show significant difference, as indicated by the C_αRMSD of 1.0 Å for all TMs. In addition, the interdomain hydrogen bonding and salt-bridge interactions are similar (hydrogen-bond number = 5.06 ± 1.64 and 5.07 ± 1.83 for the 135Psub and 135NPsub systems). Therefore, the stabilization by Glu¹³⁵ protonation may arise from the local structural variation rather than the global conformational change.

Visual inspection over the microenvironment around Glu¹³⁵ indicates that this ionizable residue is buried within an environment composed of several highly hydrophobic residues (Fig. S11, A and C), including Phe³⁵, Trp³⁸, Leu¹³¹, Leu¹³⁴, Ile³⁹², and Val³⁸⁸, most of which are conserved among various species (25). On the other hand, the carboxyl group of Glu¹³⁵ is not well solvated by water, as indicated by its relatively small solvent-accessible surface area (SASA, mean value of 12.00 Å² with a 95% confidence interval of [11.56, 12.45]) in the 135Psub system when compared with the outward-open state (mean SASA of 19.27 Å² with a 95% confidence interval of [18.97, 19.56]). Accompanying the outward-to-inward structural transition, the ionizable Glu¹³⁵ is thus gradually transferred into a relatively more hydrophobic (and less solvent-exposed) microenvironment, which spontaneously favors the protonation of this residue to an electrically neutral species, due to the desolvation penalty for charge burial. To test this hypothesis, we calculated the pK_a values for all frames in the largest cluster from the 135Psub system using PROPKA (41,42). The pK_a values of 7.49 ± 0.31 (in 645 frames) indicate that Glu¹³⁵ tends to be protonated in the inward-facing state, particularly when compared with the pK_a values of Glu¹³⁵ in the outward-open state (6.82 ± 0.47 in 1833 frames). The significant rise of pK_a (p value < 10⁻¹⁵) by 0.67 units implies that Glu¹³⁵ is more likely to bind a proton in the inward-facing state than in the outward-facing state.

Furthermore, by neglecting the structural difference between the Glu¹³⁵ protonated and deprotonated proteins and by assuming the rise of pK_a as the only contributing factor for free energy, the difference of free energy changes (ΔΔG) during structural transition between the Glu¹³⁵ protonated and deprotonated systems can be estimated from the ΔpK_a values during the structural transition, using $\Delta \Delta G=-2.303\times RT\times \Delta {\text{pK}}_{\text{a}}$ (see Fig. S12 for details), where R is the ideal gas constant, and T is the absolute temperature. As a simple estimation at 310 K, we have ΔΔG = −0.95 kcal/mol, which accounts for ∼43% of the overall ΔΔG value calculated from the along-the-pathway PMF profile (−2.19 kcal/mol). Therefore, the topological design of local hydrophobic microenvironment around the ionizable residue Glu¹³⁵ contributes >40% to the structural stabilization of the inward-facing conformation provided by the protonation of Glu¹³⁵.

The relative stability of the inward-facing conformation in the two systems may also partially arise from the distinct binding manners of the substrate fucose. In the 135Psub system, the nonpolar methyl group is favorably harbored by several hydrophobic residues (Fig. S11 B), as indicated by the mean SASA of 3.92 Å² with a 95% confidence interval of [3.60, 4.24] among all structures around the inward-facing state (State 4 on the d_IC-d_EC map). In contrast, in the 135NPsub system, the methyl group is exposed to solvent to a larger extent (Fig. S11 D), as indicated by the mean SASA of 17.97 Å² with a 95% confidence interval of [17.23, 18.71].

Helix unwinds locally in the transition state along the reaction pathway

Along the structural transition pathway, a huge free energy barrier exists between the occluded state and the inward-facing state for both systems (Fig. 4 D), which corresponds to the transition states. According to the movie constructed from frames equally distributed along the reaction pathway of the 135Psub system (Movie S3), an abnormal kink occurs in TM11 during the occluded-to-inward transition. We then extracted structures around the transition state (saddle point in Fig. 4 A) from the equilibration trajectory of the 135Psub system, clustered the frames by C_αRMSD, and compared the representative structure (snapshot closest to the center of the largest cluster) with the occluded state. Indeed, the transition state exhibits a severe distortion at the center of TM11 (Fig. S13 A). In addition, as compared with the occluded state, TM7 in the transition state significantly reorients to further close the periplasmic gate and to enlarge the cytoplasmic opening, a key step that is essential during the occluded-to-inward structural transition. The reoriented TM7 in the transition state cannot be stabilized sufficiently by the tightly packed neighboring helix TM11 unless the latter is bent in the middle. In other words, the local distortion at the center of TM11 supposedly facilitates the reorientation of TM7 and thus the occluded-to-inward structural transition. After TM7 is successfully reoriented, the TM11 restores its straight helical structure, as shown by the comparison between the transition state and the inward-facing one in Fig. S13 B.

Close inspection on the structural distortion in TM11 shows that the helical conformation from residue 393 to 399 is almost completely disrupted in the transition state. Particularly notably, there are three successive glycines (Gly³⁹³, Gly³⁹⁴, and Gly³⁹⁵) in this segment, which enable the local helix unwinding, and at least two of them are highly conserved within the same protein family (Fig. S14). In addition, the Pro³⁹⁹ that is also conserved may facilitate the helix break. According to the movie generated from the PCA analysis on all snapshots from the ABF simulation of the 135Psub system, the helical unwinding mainly occurs in the second PC mode, a degree of freedom that is nearly orthogonal to the reaction pathway (Movie S4). The continuous sampling along this degree of freedom in our free energy calculation therefore supports the validity of the transition state identified in this work (Fig. S6). After the helix unwinds in the middle, two of the three glycines are poorly hydrogen-bonded (Fig. S13 C), as suggested by 63.9, 94.1, and 63.2% of hydrogen-bonded frames for them, respectively, within all structures collected around the transition state. As a contrast, the percentages of hydrogen bonds for these residues exceed 83% in the low energy states (Table S2), especially for the Gly³⁹³, which is hydrogen-bonded in 95–98% of the frames around the outward-open and inward-facing states. Furthermore, the unpaired hydrogen-bonding donors or acceptors of these three glycines are buried in a hydrophobic environment (Fig. S13 D) composed of both nonpolar residues and membrane lipids, which explains the high free energy of the transition state.

The transport mechanism of FucP

MFS transporters are generally believed to accomplish the substrate transport using the alternating access model, which requires that the transporter alternatingly approaches the outward-open and inward-open conformations in order to continuously transport the substrate from one side of the membrane to the other side. Our simulations support this model. Starting from the outward-open crystal structure, we observed the large-scale conformational change of FucP from the outward-facing to the inward-facing conformation in the presence of substrate. Because the structural snapshots from the AT simulations have been sufficiently equilibrated, we finally obtained an inward-facing conformation with favorable stabilizing interactions. Free energy calculation suggests that this inward-facing conformation is only slightly less stable than the outward-open one (free energy difference < 1 kcal/mol) in the presence of both proton association and substrate binding, which therefore fulfills the thermodynamic requirement of the alternating access model. Moreover, two intermediate states have been observed in our simulations, which include the outward-facing partially opened conformation and the occluded one.

As a fucose/H⁺ symporter, FucP has to transport the two cargos simultaneously. Only in this way could the fucose be transported against the chemical gradient by utilizing the established cross-membrane proton gradient. How is the proton gradient effectively utilized for substrate transport? Our simulations suggest that it is accomplished through a key ionizable residue Glu¹³⁵. By changing its protonation state, Glu¹³⁵ bridges the proton association and protein conformational change, the latter of which is strongly correlated with substrate transport according to the alternating access model. The substrate transport cycle of FucP starts from an unloaded outward-open conformation and the substrate might bind before the proton association, because the neutralizing mutation E135Q devastates the substrate binding according to previous experimental studies (27). With the substrate bound, the transporter gradually closes its periplasmic entrance. However, without the association with a periplasmic proton, FucP is highly likely to stay in the outward-facing partially opened conformation. In the presence of abundant extracellular protons, the association of proton with Glu¹³⁵ reverses the relative stability of the two intermediates and therefore triggers the protein to visit the occluded state. Although the presence of a large free energy barrier prevents quick ongoing structural transition, the protonation of Glu¹³⁵ effectively favors the reaction through stabilizing the inward-facing conformation by >2 kcal/mol. Moreover, in the Glu¹³⁵ protonated protein, the overall free energy barrier along the reaction pathway is reduced by 1.7 kcal/mol, a number that corresponds to ∼16-fold increase in the substrate transport rate at the physiological temperature according to Eyring’s transition-state theory (50). Therefore, the alterations in thermodynamics and kinetics jointly explain the promotion of the outward-to-inward transition by the protonation of Glu¹³⁵. From the inward-facing conformation, the protein may need to further open its cytoplasmic entrance so as to release both substrate and proton into cytosol, after which the protein naturally returns to the unloaded starting conformation to complete a cycle of conformational change. Unfortunately, the complete opening of the cytoplasmic entrance from the inward-facing state was not observed in our simulations, possibly because of the insufficient sampling in the AT simulation or because of the presence of an unidentified reaction pathway. Future work should be done to catch the complete inward-open conformation and to estimate the free energy profile more reliably. In addition, the inward-to-outward transition of the unloaded transporter should also be explored in future studies. Nevertheless, the coupling of proton binding and conformational change in FucP reported here may shed light on the mechanistic studies of other MFS transporters that utilize the gradient of protons or ions to transport substrates.

Author Contributions

Y.L. and H.G. designed research; Y.L. performed research; Y.L., M.K., and H.G. analyzed data; and Y.L. and H.G. wrote the article.

Editor: Carmen Domene.

Supporting Material

Movie S1. 135Psub structural transition

The conformational change of the 135Psub system in the AT simulation. (Blue and yellow) NTD and CTD, respectively; (red) fucose.

Movie S2. 135NPsub structural transition

The conformational change of the 135NPsub system in the AT simulation. (Blue and yellow) NTD and CTD, respectively; (red) fucose.

Movie S3. Along-the-pathway conformational change

The smoothed conformational change of FucP along the pathway of structural transition. To generate the movie, the reaction pathway was equally partitioned and the representative frame in each interval was taken from the ABF simulations.

Movie S4. TM11 unwinding in PC2

The local unwinding of TM11 is mainly reflected in the second PC mode identified in the PCA analysis. For clarity, only the TM helices are shown (ribbon representation). (Red) TM11; (spheres) C_α atoms of Gly³⁹³–³⁹⁵. (Blue and yellow) NTD and CTD, respectively.