Anomalous behavior of water inside the SecY translocon

Significance

α-Helical membrane proteins are assembled cotranslationally into cell membranes with the aid of the heterotrimeric membrane-embedded translocon complex. The structures of several translocons reveal that the interior has an hourglass shape filled with water. Because nascent peptide chains emerging from the ribosome are thought to pass through the translocon during secretion or membrane insertion, the translocon is often referred to as a protein-conducting channel. We show by all-atom molecular-dynamics simulations that the water molecules within a translocon do not behave as in a bulk phase, which raises fundamental questions not only about the current conceptual model of translocon-assisted insertion of membrane proteins but also about the physical principles of partitioning of solutes between bulk and confined water.

Abstract

The heterotrimeric SecY translocon complex is required for the cotranslational assembly of membrane proteins in bacteria and archaea. The insertion of transmembrane (TM) segments during nascent-chain passage through the translocon is generally viewed as a simple partitioning process between the water-filled translocon and membrane lipid bilayer, suggesting that partitioning is driven by the hydrophobic effect. Indeed, the apparent free energy of partitioning of unnatural aliphatic amino acids on TM segments is proportional to accessible surface area, which is a hallmark of the hydrophobic effect [Öjemalm K, et al. (2011) Proc Natl Acad Sci USA 108(31):E359–E364]. However, the apparent partitioning solvation parameter is less than one-half the value expected for simple bulk partitioning, suggesting that the water in the translocon departs from bulk behavior. To examine the state of water in a SecY translocon complex embedded in a lipid bilayer, we carried out all-atom molecular-dynamics simulations of the Pyrococcus furiosus SecYE, which was determined to be in a “primed” open state [Egea PF, Stroud RM (2010) Proc Natl Acad Sci USA 107(40):17182–17187]. Remarkably, SecYE remained in this state throughout our 450-ns simulation. Water molecules within SecY exhibited anomalous diffusion, had highly retarded rotational dynamics, and aligned their dipoles along the SecY transmembrane axis. The translocon is therefore not a simple water-filled pore, which raises the question of how anomalous water behavior affects the mechanism of translocon function and, more generally, the partitioning of hydrophobic molecules. Because large water-filled cavities are found in many membrane proteins, our findings may have broader implications.

The heterotrimeric SecY translocon complex (SecYEG in bacteria, SecYEβ in archaea, Sec61αβγ in eukaryotes) is required for the cotranslational assembly of membrane proteins and the secretion of soluble proteins (1–3). The SecY subunit (Fig. 1A) has 10 transmembrane helices comprised of two five-helix domains related by pseudo-twofold symmetry around an axis parallel to the membrane (4–8). These helices form an hourglass-shaped water-filled pore that spans the membrane. The so-called hydrophobic pore ring (HR) comprised of six hydrophobic residues forms the narrowest part of the hourglass, located near the bilayer center. Sitting just above the ring on the extracellular side is a small, distorted helix [transmembrane 2a (TM2a)], called the plug domain that is believed to impede the passage of water and solutes across the membrane. Access to the membrane from the water-filled hourglass-shaped interior of SecY is provided by the so-called lateral gate, formed by helices TM2b and TM7 (Fig. 1A). SecG is not required for function, but SecE is indispensable. Experimental (9–11) and computational studies (12–15) emphasize the importance of interactions of the gate helices with nascent-chain segments during TM helix insertion.

Fig. 1.

Structure of P. furiosus SecYE and cartoon representation of the translocon-to-lipid bilayer partitioning of a TM helix. (A) Structure of SecYE in a lipid bilayer. The lipid headgroups are shown in ice-blue van der Waals representation. The 10 SecY helices are represented in gray cartoon format, except for TM2b (magenta) and TM7 (cyan). The SecE helices are colored green. Hydrophobic ring (HR) residues are drawn as yellow bonds. The plug domain TM2a is colored orange. (B) The apparent free energy of partitioning ΔG(z) of an alkyl sidechain (orange) with accessible surface area A_acc depends upon position z within the TM segment (solid red curve), because the atomic solvation parameter σ apparently depends upon position. Our results suggest an opposite behavior (dashed orange curve).

The general belief is that nascent chains pass through the translocon via the normally closed HR, which must open in response to ribosome docking and nascent-chain elongation. When a TM segment passes through SecY, it is believed that the segment is shunted into the membrane through a simple partitioning process between water-filled SecY and the lipid environment (Fig. 1B). Hessa et al. (16, 17) characterized the translocon/membrane insertion energetics and established that the membrane insertion propensity depends on both the hydrophobicity and position of each residue within model TM segments (16, 17). Because the resulting “biological” hydrophobicity scale (16, 17) correlates strongly with physical hydrophobicity scales, translocon/membrane partitioning is believed to be similar to water/membrane partitioning of peptides (18), which is driven by the hydrophobic effect (19).

The hallmark of hydrophobic partitioning is that the favorable partitioning free energy is linearly proportional to the solute nonpolar accessible surface area and is described by the atomic solvation parameter σ (20). For bulk water-to-hydrocarbon partitioning, the solvation parameter typically has a value of about −23 cal⋅mol⁻¹⋅Å⁻² (21). Öjemalm et al. (22) found that the apparent translocon-to-membrane free energy of insertion of nonproteinogenic amino acids at a particular position in the TM segment varied linearly with σ, as expected for hydrophobic-driven partitioning. However, σ was found to be less favorable than that for bulk water/hydrocarbon partitioning and to depend on position within the model TM segment (Fig. 1B). Near the ends of the segments, σ ≈ −6 cal⋅mol⁻¹⋅Å⁻², whereas in the center of the segment σ ≈ −10 cal⋅mol⁻¹⋅Å⁻² (22). These low values of σ caused us to examine the physical behavior of water within the translocon. Specifically, could the water properties within the translocon explain the magnitudes and position dependence of the translocon/bilayer partitioning solvation parameters? To address this question, we turned to atomistic molecular-dynamics (MD) simulations.

From a number of crystallographic SecY structures (4–8), we chose the SecYE structure from Pyrococcus furiosus determined by Egea and Stroud (8), because it appears to be in a nearly open (“primed”) state as judged by the separation of the gate helices. We thought at the outset of our simulations that SecYE might close. However, SecYE remained stably open, which allowed close examination of the waters inside SecY. We found that waters deep within the translocon diffuse anomalously, have slow rotational dynamics, and have their dipoles aligned along the SecY axis. These properties indicate that translocons are not simple water-filled pores, which raises fundamental questions about the nature of translocon/bilayer partitioning of TM segments, and more generally about solute partitioning into compartments of restrained water molecules.

Results

SecYE Remained Stably Open During the Simulation.

We began our study with a 0.45-μs simulation (Sim1) of SecYE embedded in a palmitoyloleoylphosphatidylcholine (POPC) bilayer in excess water using NPT (constant number of particles N, pressure P, and temperature T) conditions (Methods). Egea and Stroud (8) reported the P. furiosus SecYE translocon structure to be in a primed open state based upon the dimensions of the HR. We monitored the HR radius throughout the simulation (SI Appendix, Fig. S1) and found it to be quite stable with a mean radius of 6.8 ±0.3 Å (SD) (see Fig. 3D). The same measurement in the closed Methanococcus jannaschii SecYEβ yielded 4.6 ± 0.1 Å, as observed by Egea and Stroud (8) (SI Appendix, Fig. S2). Consistent with the stability of the ring, the SecY TM region was rather rigid (C_α rmsd < 2 Å; SI Appendix, Fig. S4A), confirming that SecY did not undergo any major conformational changes during the simulation. We confirmed the stability of the SecY structure by carrying out two additional NPT simulations of 130 ns, described in Supporting Information (SI Appendix, Figs. S4 and S5).

Fig. 3.

Fluctuations of water occupancy in the HR region, due in part to incursions of lipid acyl chains. (A) Water occupancy calculated over the 0.45-μs simulation in the region enclosing the HR residues. Black dashed lines indicate the times of the snapshots in B and C; red dashed vertical line defines the time t = 0.15 μs at which the lipid acyl chains start exploring the HR region (SI Appendix, Fig. S3); red horizontal lines indicate time ranges of NVE simulations. (B and C) Representative snapshots taken at t = 0.06 μs (B) and t = 0.28 μs (C). The color scheme is the same as in Fig. 1, except the plug domain TM2a (orange, new cartoon format). Water is represented using both van der Waals and surface representations. Lipids in front of the lateral gate are shown in blue van der Waals format. The gray dashed lines indicate the region explored by the HR residues along the membrane normal (z axis). (D) Time evolution of the HR radius of SecY. The average value is 6.8 ± 0.3 Å.

Water Distribution Within SecYE Conforms to the Hourglass Shape of the Interior.

To examine the water distribution within SecY, we defined a 40 × 40 × 100-ų square prism (“the prism”) enclosing SecY and centered at the simulation cell origin (Fig. 2B). The placement resulted in the HR ring being located between z = ±4 Å (SI Appendix, Fig. S1B). We then determined the time-averaged number of water molecules within the prism in consecutive slices of 1-Å thickness along the z axis (Fig. 2C). The average total number of water molecules within the hourglass between z = ± 20 Å was about 430.

Fig. 2.

Water distribution within the SecYE P. furiosus translocon. (A) Side-view snapshot from the cell of the P. furiosus SecYE translocon embedded in a POPC lipid bilayer. The SecY structure and coloring is the same as in Fig. 1A. Water molecules inside the 40 × 40 × 100-ų square prism are shown in red (oxygen) and white (hydrogen) van der Waals representation, and those outside in a dimmed surface representation. Periplasmic and cytoplasmic sides of the membrane are indicated. (B) Representation of the 20 5-Å-thick slabs (black lines) parallel to the membrane plane used to subdivide the prism for studying water as a function of position along the membrane normal. Coloring is as in A. (C) Logarithmic representation of water distribution along the membrane normal from the first 0.15 μs of the 0.45-μs Sim1 determined from the time-averaged number of water molecules at 1-Å intervals. The dashed horizontal lines indicate the outermost slabs and the HR location.

Dewetting phenomena have been observed in MD simulations of hydrophobic protein channels and gates (23–26). We looked for dewetting in the HR region by monitoring the water occupancy within the volume enclosed by the HR sidechains (Fig. 3A and SI Appendix, Fig. S5). We observed dramatic fluctuations in the number of waters due to lipid acyl chains exploring SecY (Fig. 3 B and C, and SI Appendix, Fig. S3), but there was no evidence of persistent dewetting (Movie S1). The lipids located in front of the gate generally confined the water molecules within SecY (Fig. 3B and Movie S2), and they transiently explored the SecY interior, perturbing water passage (Fig. 3C, SI Appendix, Fig. S3, and Movies S1 and S3). These perturbations took place only after the first 0.15 μs of Sim1 (Fig. 3A). To understand how the acyl chains perturbed water within SecY, we examined the time-averaged number of acyl chain carbon atoms inside the prism (SI Appendix, Fig. S3A). The major acyl chain incursions occur between z = ± 10 Å, but the acyl chain carbon atoms were not able to explore the HR region (SI Appendix, Fig. S3B) during the first 0.15 μs of Sim1, and only to a limited extent after that time. Because the TM helix insertion process involves direct interactions among lipids, water, translocon, and incoming peptides (11, 16), we suppose that these acyl chains—noted in earlier simulations (27, 28)—are likely to play a role in TM helix insertion. However, we do not know whether the lipids we observed are physiologically significant. Similar incursions were seen in Sim2 and Sim3 (SI Appendix, Fig. S5).

Water Dynamics Within the Hourglass Deviates Dramatically from Bulk Behavior.

For the analysis of the NPT simulations, we stored the system coordinates every 10 ps, but this time interval is too long to reveal the details of water dynamics that occur on the picosecond timescale (29, 30). Therefore, we carried out 40 NVE (constant number of particles N, volume V, and energy E) simulations of 50-ps length, saving configurations every 5 fs (Methods). Each of the simulations was started from independent initial structures selected from five different time intervals (red horizontal lines, Fig. 3A) of Sim1. We determined the translational diffusion and the rotational dynamics of water in the prism enclosing SecY (Fig. 3B). We began by examining the first 0.15 μs of Sim1 (dashed vertical red line, Fig. 3A), because the water-conducting states within HR were unperturbed by acyl chain incursions (SI Appendix, Fig. S3B).

We characterized the translational motion of waters by calculating the dependence of mean squared displacements (MSDs) on time (SI Appendix, Fig. S6A). In general, after a few picoseconds, the MSD became proportional to t^α (29). Linear dependence of the MSD on time (α = 1) is a signature of Brownian diffusive motion as observed for bulk water; α < 1 is a signature of anomalous “subdiffusion,” which is characteristic of confined water and protein hydration water. The MSD of water molecules in the long-time regime (t > 2 ps) behaved differently in each of the 5-Å-thick slabs, as characterized by α (Fig. 4A, solid black circles). Approaching the SecY center, the α decreased from 1 to 0.53, which is similar to values reported for protein hydration water (30–35). The diffusion of water deep within the translocon is therefore anomalous with respect to bulk behavior.

Fig. 4.

NVE simulations reveal anomalous water motions inside SecY. A and B show data for t < 0.15 μs; C and D show data for t > 0.15 μs. Black circles (SecY) show values from unrestrained simulations; green diamonds, those from SecY-fix; blue diamonds, those from SecY-vdw. (A and C) Variation of α exponent variation along the membrane normal estimated from the NVE simulations. (B and D) The variation of the rotational relaxation time τ_μ along the membrane normal estimated from the NVE simulations.

Water dynamics can also be assessed using single-molecule dipole autocorrelation (SDAC) functions C_μ(t), whose decays are determined by local librational and rotational motions of water (Methods). We defined the rotational characteristic relaxation time τ_μ as the time required for C_μ(t) to decay to 1/e to analyze the SDAC functions consistently. The functions exhibited very different trends depending on the location of water within SecY (SI Appendix, Fig. S6B). Within the 5-Å-thick slabs at the extreme ends of the prism (Fig. 4), a relaxation time τ_μ ∼ 1 ps was observed, as expected for bulk water (36) (Fig. 4B, SecY, solid black circles). Inside SecY, however, τ_μ progressively increased to 20–30 ps, due to water–SecY interactions that restricted the reorientational motion of water molecules. Near the center of SecY, water relaxation exceeded the observation time window so that only a lower limit could be determined.

We next performed the same analyses on trajectories obtained from the NVE simulations run at times t > 0.15 μs, when lipid acyl chains were exploring the interior of SecY. The results revealed an increase in the perturbation of water translational dynamics: α decreased from 1 to 0.40 as the SecY center was approached (Fig. 4C). In contrast, lipid incursions influenced water rotational dynamics only weakly (Fig. 4D). Approaching the SecY center, the relaxation times are about the same as when water is unperturbed by the lipid acyl chains.

Electrostatic and Confinement Effects Determine Dynamics of Water.

What features of SecY cause anomalous water behavior? One possibility is restriction of water movement due to the limited volume of the hourglass (confinement). Another possibility is electrostatic interactions between water and the translocon. We designed two additional 40 NVE setups (Methods) using the same configurations selected from Sim1. In the first setup, the coordinates of SecY were fixed (SecY-fix; SI Appendix, Figs. S6 C and D and S7 C and D). This approach allowed us to determine the effect of SecY thermal motion on water dynamics. In the second setup (SecY-vdw), in addition to “freezing” SecY, we switched off all SecYE charges, including partial charges (SI Appendix, Figs. S6 E and F and S7 E and F) to separate confinement from electrostatic effects. For t < 0.15 μs, water exhibited essentially the same translational dynamics regardless of whether SecY was mobile or frozen (Fig. 4A, solid black circles and solid green diamonds, respectively), meaning that the thermal motion of SecYE had little effect on diffusion. However, after switching off electrostatic interactions, water still exhibited anomalous diffusion, but the value of α at the SecY center was 0.72 rather than 0.53 (SecY-vdw in Fig. 4A). For t > 0.15 μs, when the lipids make incursions inside SecY, α reached 0.63 at the SecY center (SecY-vdw in Fig. 4C, blue diamonds) instead of 0.40. These findings indicate that both confinement and the SecY electrostatics make water translational dynamics depart from bulk diffusion.

Water rotational dynamics slowed upon freezing SecY (Fig. 4 B and D; compare black symbols and green diamonds) irrespective of the incursions of the lipid acyl chains inside the SecY pore. We speculate that restraining SecY thermal motions perturbs the rearrangement rate of the SecY hydrogen bond network (37) as well as that of the SecY–water network, thus affecting the water rotational dynamics observed when SecY is mobile (Fig. 4 B and D; solid black circles labeled SecY). Interestingly, once the SecY charges are switched off, water characteristic rotational relaxation times revert to bulk values, typically ∼1 ps, throughout the translocon (Fig. 4 B–D, solid blue diamonds). Despite the bulk-like relaxation times, the SecY-vdw SDAC functions of water molecules located inside SecY differ considerably from those of bulk water (SI Appendix, Figs. S6F and S7F), which indicates that water inside SecY in the absence of the electrostatic field behaves differently from bulk water.

Water Molecules Are Highly Oriented Inside the SecYE Translocon.

The simplest way to examine the effect of the SecY electrostatics on water structure and the interpretation of the relaxation time τ_μ in the absence of charges is to determine the average orientation of water dipoles with respect to the SecY z axis using the order parameter P_d (z) = <cosθ>(z) (Methods). A random distribution of dipole moments corresponds to P_d (z) = 0. We first analyzed Sim1 and calculated the distribution of the water dipole orientation over three time windows: 0–0.15 μs, 0.15–0.3 μs, and 0.3–0.45 μs. Our analysis of P_d is displayed in Fig. 5 along with a representative snapshot in which water oxygen atoms are colored according to the dipole orientation calculated between 0 and 0.15 μs. These results show that water exhibits nearly the same behavior over the whole length of Sim1. Outside SecY, water dipoles do not show any net orientational preference. Within SecY P_d varies between −0.2 and 0.6, with the strongest orientation <cosθ> = 0.6 occurring near the hydrophobic ring. This result reflects a very high degree of alignment, and explains the trends observed for the SecY SDAC functions in this region (SI Appendix, Figs. S6B and S7B). The lack of C_μ(t) decay within the time window studied is a signature of an almost fixed dipole direction.

Fig. 5.

Water dipoles are aligned near the center of SecY. (A) Water dipole orientations were calculated from 0 to 0.15 μs (black symbols), from 0.15 to 0.30 μs (orange symbols), and from 0.30 to 0.45 μs (red symbols) of the NPT simulation Sim1 and are shown as a function of position along the membrane normal. (B) Representative snapshot of the orientation of the water dipoles within the SecY channel. Water oxygen atoms are drawn as beads and colored according with the value of the order parameter calculated between 0 and 0.15 μs as indicated by the colored bar. SecY is represented using the color scheme of Fig. 1.

The dipole alignment observed is comparable with that exhibited by water in the aquaporin channel (38) or in a mechanosensitive channel (39). To clarify whether or not the alignment was due to electrostatics or confinement, we calculated the average orientation of the water dipoles for all of the 120 NVE simulations, including SecY-fix and SecY-vdw (SI Appendix, Fig. S8 A and C). When the SecY charges were turned off (SecY-vdw, blue symbols), the water dipoles were still aligned to the membrane normal, but less than observed in the SecY simulations (SI Appendix, Fig. S8 A and C, compare blue and black symbols). This observation agrees with the distinct, non–bulk-like evolution of the SecY-vdw SDAC functions (SI Appendix, Figs. S6F and S7F), despite the significantly decreased decay times. We thus conclude that both the SecY electrostatics and the confinement affect the water behavior. Moreover, the presence of lipid acyl chains within the SecY pore further affects the intensity and direction of the dipole alignment (SI Appendix, Fig. S8 C and D).

Discussion

Our simulations bear strongly on the interplay between the translocon and the lipids. During our 0.45-μs simulation, the hydrophobic pore ring remained stably open independent of lipid intrusions (Fig. 3D). These results suggest that the primed open configuration of SecYE may represent a stable conformational state on the timescale of our simulation. The involvement of the lipids in the nascent-chain integration process has been suggested in previous experimental (11, 40) and computational (27, 28) studies. Our simulations provide evidence for the accessibility of the SecY pore to lipid chains of the surrounding bilayer, supporting the possibility of direct peptide-lipid interactions.

The main finding of our simulations is that water inside SecY does not behave as in bulk: translational diffusion deviates markedly from Brownian motion (Fig. 4 A and C) and the rotational dynamics show strong retardation (Fig. 4 B and D). These features are typical for water in confined environments and at hydrated protein surfaces (32, 33, 41, 42). Moreover, and perhaps more important for the thermodynamics of partitioning TM segments into the membrane, we observe that water dipoles in the HR region (z = ± 4 Å; SI Appendix, Fig. S1B) are oriented preferentially parallel to the pore axis pointing toward the exterior (Fig. 5). The inhomogeneous nonbulk properties of water within SecY suggest that translocon/membrane partitioning cannot be compared directly to partitioning between bulk aqueous and lipid phases. Contrary to expectations formed in the days when the structures of only a few membrane proteins were known, we now know that the interiors of membrane proteins (43–47), and even soluble proteins (48, 49), can contain considerable amounts of water, which raises the question of how restricted waters might participate in or affect protein function.

Our results reveal the importance of the SecY electrostatics, which might explain, for example, why flanking charges in model TM segments affect insertion by the Sec61 translocon (50). Moreover, the dipole alignment of water molecules within SecY might be a signature of another potential role of water in the membrane insertion process. For instance, it has been shown that water can facilitate or screen the interactions between lipids and a peptide located inside SecY (15). Furthermore, the dipole alignment could be crucial for the interaction of the positively charged N terminus of a signal sequence with the translocon and consequentially for membrane protein topology. The presence of a highly polarizing field will likely affect helix insertion. If the idea that incoming peptides pass directly through the SecY hourglass is correct, they would likely replace a substantial number of water molecules in the translocon interior. Hence, they would be exposed to an electrostatic field comparable to that experienced by the water molecules in our simulations. A dipole orientation that mimics these aligned water dipoles could therefore have a substantial stabilizing effect.

Favorable water-to-hydrocarbon partitioning free energies of hydrophobic moieties observed in bulk partitioning arise, in part, from the release of oriented water molecules at hydrophobic residue surfaces (51). Given a hydrophobic α-helix inside the SecY translocon that simply partitions parallel to the membrane plane into the bilayer (Fig. 1B), one would expect the solvation parameter σ for hydrophobic amino acids at the ends of the helix to be similar to bulk values, because water at the extreme ends of the translocon is bulk-like. Öjemalm et al. (22) observed, however, σ at the helix ends was −6 kcal⋅mol⁻¹⋅Å⁻², which is about one-fourth of the value expected for partitioning from bulk water. Given that water is restrained at the center of the translocon, what value of σ would one expect for amino acids at the center of a hydrophobic α-helix? To answer that question, other questions must be answered. Is water present inside SecY when a helix is present? If water is present and restrained due to SecY electrostatics and confinement, how is σ affected? If water is absent, what is the meaning of σ? One scenario might be that amino acids located in the center of a helix in the translocon have neighboring oriented water molecules that remain ordered even after the helix moves into the bilayer. In that case, the apparent solvation parameter at the helix center should be less favorable than at the helix ends. Thus, in experiments such as those of Öjemalm et al. (22), one might expect solvation parameters to be least favorable near the helix center and most favorable near the helix ends (dashed orange curve, Fig. 1B), which is exactly opposite to the results of Öjemalm et al. (solid red line, Fig. 1B). This contradiction raises fundamental questions about the translocon/membrane partitioning process of TM helices. Cymer et al. (52) have suggested that nascent TM helices may preferentially interact with the gate helices and the membrane interface (53, 54) without first passing through the SecY hourglass.

Finally, the crucial question our results raise is whether the concept of hydrophobic effect-driven partitioning even applies when a nonpolar solute moves from a restrained water region into the water-free bilayer interior. Or more simply and fundamentally, what is the partitioning free energy of solutes between water in bulk and water in restraining confined spaces? Although much theoretical and experimental work has been devoted to describing the properties of water in bulk (55) and in confined spaces (56) such as inverted micelles (57), this fundamental question remains unanswered as far as we can establish.

Methods

MD Simulations in the NPT Ensemble.

The atomic coordinates of the SecYE translocon were extracted from the Pyrococcus furiosus crystal structure (8) [Protein Data Bank (PDB) ID code 3MP7]. From this structure, we set up and carried out three independent NPT simulations: Sim1, Sim2, and Sim3 (SI Appendix, Table S1). In all cases, we embedded SecYE in a POPC bilayer formed by 600 lipids (300 molecules each leaflet). The protocols we used to model residues missing in the original structure, to embed the completed SecYE structure in the POPC lipid bilayer, and to minimize and equilibrate the systems are reported in Supporting Information. We also carried out MD simulations using the M. jannaschii SecYEβ translocon crystal structure (4) (PDB ID code 1RHZ). The same protocols were followed as for P. furiosus SecYE.

We used NAMD (58, 59), version 2.9, with the CHARMM36 (60) force field for the lipids and the CHARMM22 force field with the CMAP correction for the protein and ions (61, 62). The TIP3P model (63) was used for water molecules. The temperature was kept constant at 300 K using a Langevin dynamics scheme, and the pressure was maintained constant at 1 atm using anisotropic coupling in conjunction with Nosé–Hoover–Langevin piston algorithm (64, 65). Periodic boundary conditions were applied in three dimensions. The electrostatic interactions were computed by means of the smooth particle-mesh Ewald summation method (66, 67) and the short-range real-space interactions were cut off at 12 Å using a switching function between 10 and 12 Å. The equations of motion were integrated with a time step of 4 fs for the long-range electrostatic forces, 2 fs for the short-range nonbonded forces, and 1 fs for the bonded forces by means of a multiple-time step algorithm (68). The SHAKE (69) algorithm was used to constrain the length of the bonds involving hydrogen atoms. Coordinates were saved every 10 ps.

MD Simulations in the NVE Ensemble from Sim1.

To characterize water properties within SecY and to investigate how the SecY pore affects water dynamics, we performed five sets of eight NVE simulations, i.e., 40 NVE simulations in total. We selected the eight independent configurations every 2 ns from the following five time ranges of Sim1: 2–16, 60–74, 104–118, 302–316, and 402–416 ns. We used these 40 structures as starting configurations for running NVE simulations of 50-ps length. To have high resolution in time, we collected coordinates every 5 fs.

In addition to these five sets of eight NVE simulations in which the protein complex was unrestrained (labeled SecY in Fig. 4), we used the same starting structures to carry out two more groups of five sets of eight NVE simulations, applying restraints as follows. In the first group, we applied harmonic restraints to freeze SecY (labeled SecY-fix in Fig. 4), and in the second, in addition to freezing SecY, we switched the SecY–water electrostatic interactions off by setting the protein atom charges to zero (labeled SecY-vdw in Fig. 4). These three groups of 40 NVE simulations (SecY, SecY-fix, and SecY-vdw) enabled us to examine how the cavity shape and the electrostatics of SecY affect individually the dynamics of water. In total, we generated and analyzed 120 NVE simulations, summarized in SI Appendix, Table S2.

Water Anomalous Dynamics.

We investigated water dynamics in a selected region of the simulation cell characterized by a volume of 40 × 40 × 100-ų encompassing SecY (Fig. 2B). We subsequently divided this square prism into 20 5-Å-thick slabs parallel to the membrane (Fig. 2C). For each slab, we first computed and analyzed the MSDs, which characterize the translational diffusion of water molecules. MSDs are defined as follows:

$$MSD\left(t\right)=\langle {\left|{\overline{r}}_i\left(t\right)-{\overline{r}}_i\left(0\right)\right|}^2\rangle ,$$

where ${\overline{r}}_i\left(t\right)$ denotes the position of the particle i at time t, and the brackets denote an average over molecules and time origins. In SI Appendix, Figs. S6 A, C, and E, and S7 A, C, and E, we present the MSD of the center of mass of water molecules computed for the SecY, SecY-fix, and SecY-vdw for each slab and for times t < 0.15 μs and t > 0.15 μs. In the long-time regime, the time dependence of the MSD can be described with a power law MSD(t) = kt^α, and the α exponent gives information on the diffusive character of water motion. We fitted the MSD to the power-law function for t > 2 ps and represented the α values in Fig. 4. Further details of the protocol used for the data analysis are reported in Supporting Information.

Subsequently, we calculated the SDAC functions C_μ(t), which signify the rotational motion of water molecules and are defined as follows:

$$C_{\mu }\left(t\right)=\frac{\langle \widehat{\mu }\left(t\right)\cdot \widehat{\mu }\left(0\right)\rangle }{\langle \widehat{\mu }\left(0\right)\cdot \widehat{\mu }\left(0\right)\rangle },$$

where $\widehat{\mu }\left(t\right)$ is the unit vector of the water dipole at time t. In SI Appendix, Figs. S6 B, D, and F, and S7 B, D, and F, we have plotted the SDAC functions computed for SecY, SecY-fix, and SecY-vdw for each slab for the time windows t < 0.15 μs and t > 0.15 μs. We analyzed consistently all of the 120 NVE simulations and calculated the characteristic relaxation time τ_μ as the time at which C_μ(t) decays to 1/e, that is, C_μ(τ_μ) = 1/e (SI Appendix, Figs. S6 and S7, dashed lines). The τ_μ values are reported in Fig. 3. Further details of the data analysis procedure are reported in Supporting Information.

Finally, we examined the water dipole orientation within the SecY channel in terms of the water orientational order parameter defined as follows:

$$P_d\left(z\right)=\langle \cos \theta \rangle \left(z\right),$$

where θ is the angle between the water dipole and the membrane normal. The θ value varies between 0° and 180°, hence cosθ varies between −1 and 1. We calculated the values of cosθ between 0 and 0.15 μs, 0.15 and 0.3 μs, and 0.3 and 0.45 μs of Sim1 (Fig. 5), and over all of the NVE simulations (SecY, SecY-fix, and SecY-vdw) (SI Appendix, Fig. S8 A and C). The procedure we followed is described in detail in Supporting Information.