Molecular Dynamics Simulations of Hemolytic Peptide δ-Lysin Interacting with a POPC Lipid Bilayer

Abstract

The binding interaction between a hemolytic peptide ${\delta}$-lysin and a zwitterionic lipid bilayer POPC was investigated through a series of molecular dynamics (MD) simulations. ${\delta}$-Lysin is a 26-residue, amphipathic, ${\alpha}$-helical peptide toxin secreted by Staphylococcus aureus. Unlike typical antimicrobial peptides, ${\delta}$-lysin has no net charge and it is often found in aggregated forms in solution even at low concentration. Our study showed that only the monomer, not dimer, inserts into the bilayer interior. The monomer is preferentially attracted toward the membrane with its hydrophilic side facing the bilayer surface. However, peptide insertion requires the opposite orientation where the hydrophobic side of peptide points toward the membrane interior. Such orientation allows the charged residues, Lys and Asp, to have stable salt bridges with the lipid head-group while the hydrophobic residues are buried deeper in the hydrophobic lipid interior. Our simulations suggest that breaking these salt bridges is the key step for the monomer to be fully inserted into the center of lipid bilayer and, possibly, to translocate across the membrane.

Introduction

Antimicrobial peptides (AMPs)1-4 are short polypeptides, between 14 and 40 residues in length, produced by wide variety of organisms, both prokaryotic and eukaryotic, as a part of their natural defense mechanisms. These peptides from seemingly unrelated organisms share an important common feature: many of them form an amphipathic α- helix upon binding to a lipid membrane. Interestingly, there is no receptor on the target cell that can recognize and distinguish these peptides.5,6 This is important in the development of antibiotics based on AMPs, which has been pursued by the pharmaceutical industry for the past a few decades. While bacteria can easily become resistant to a given antibiotic, it is harder to develop resistance to antimicrobial peptides, as it would require bacteria to change the composition or packing of their lipids. Because of the diversity in peptide sequence and the lack of unique epitopes, the origin of target cell specificity remains elusive, not to mention the actual cytolytic mechanism that kills microbes.7-10 In particular, one of the outstanding questions is whether a specific residue sequence and/or composition are tied to a particular membrane disruption mechanism.

While AMPs, which in general have a net positive charge, work specifically against bacterial membrane, other cytolytic peptides, such as melittin and mastoparan secreted by bees and wasps, do not discriminate different cell types.11,12 Cytolytic peptides, however, are not limited to eukaryotic organisms. For example, δ-lylin is a 26-residue hemolytic peptide (formyl- MAQDIISTIGDLVKWIIDTVNKFTKK) secreted by Staphylococcus aureus, a Gram-positive bacterium.13-17 δ- Lysin possesses a couple of unique properties that distinguish itself from the rest. First, it preferentially lyses eukaryotic cells with little antimicrobial activity. Second, with four Lys, three Asp, and a negative C-terminus, δ-lysin has no net charge in neutral pH. It is often believed that the positive charge on AMPs is the key for the cell specificity since bacterial membranes are negatively charged.2,18,19 However, positively charged melittin lyses red blood cells with neutral membrane as well as bacterial cells.2 Furthermore, the effectiveness of positively charged magainin varies widely against different types of negatively charged membranes.20 Thus, the net charge on peptides cannot be the sole source of their specificity against different membranes and its effect is convoluted with other factors such as sequence and charge distribution. In this regard, studying the action of δ-lysin against neutral membrane has an advantage because the net charge is not a major factor to determine the specificity and the mechanism of membrane disruption, which simplifies the overall picture.

An earlier study of δ-lysin by Mellor et al.21 concluded based on the observation of voltage-dependent channel formation that vertically inserted helical δ-lysin hexamer form a stable pore, supporting the so-called barrel-stave model7 of channel formation. However, a fluorescence study by Talbot and coworkers22 indicated that the primary species on the membrane interface is δ-lysin dimer. More recently, through kinetic analyses of their dye efflux studies, Almeida and coworders10,23,24 suggested that δ-lysin do not form pores after all. Rather, they translocate the membrane as oligomers, possibly as trimers, in parallel orientation with respect to the membrane surface. However, the detailed mechanism of dimer and trimer formation on the surface of membrane is not clear, not to mention the translocation of oligomers across the membrane.

As implied by the conflicting reports on δ-lysin mentioned above, the action of δ-lysin against lipid bilayer has not been fully understood yet. Almeida et al.24 reported that the binding of δ-lysin on neutral membrane POPC is much better than what is anticipated from the well-known interfacial scale of Wimley and White.25 They suggested that intramolecular salt bridges between Lys and Asp are the source of discrepancy between the experimentally measured binding free energy and the prediction by the Wimley-White scale. Moreover, it was proposed that intermolecular salt bridges are necessary to form dimer and trimer for the peptide translocation.10 As an initial attempt to understand the nature of interaction between δ-lysin and neutral membrane, we performed a series of molecular dynamics simulations of δ-lysin monomer and dimer interacting with zwitterionic model bilayer POPC (1-palmitoyl-2-oleoyl-sn-glycero-2-phosphocholine). Molecular dynamics (MD) simulations have been proven as a valuable tool to provide atomistic details, both structural and dynamical information, on the protein-membrane systems. In the present work, association of δ-lysin with a POPC bilayer is investigated through careful analyses of a set of MD trajectories, starting with various peptide orientations and conformations. In particular, we examined how the key residues in δ-lysin, such as Lys and Asp, play a role in peptide-membrane and peptide-peptide interactions. The simulations also revealed how the membrane is locally disturbed by the peptides, which affects the mechanism of peptide insertion into the membrane core.

 

Computational Methods

All molecular dynamics simulations (MD) reported in this work were carried out with the GROMACS26 package. The peptides were described by gromacs force field (ffgmx) whereas a POPC bilayer was modeled by the lipid force field developed by Berger et al.27 A bilayer structure made of 128 POPC molecules was obtained from Tieleman28 and equilibrated before use. All simulations were conducted under NPT condition (323 K and 1 atm) with periodic boundary condition. Temperature was maintained through Nos-Hoover thermostat with 0.1 ps coupling constant whereas pressure was controlled by Parrinello-Raman method with 2.0 ps coupling constant. The van der Waals interaction was cut-off at 10.0 Å and a real-space cut-off at 10.0 Å was used in combination the particle-mesh Ewald (PME) method for the electrostatic interaction. A time step of 2.0 fs was used and the neighbor list was updated in every 10 steps. All bond lengths were constrained at their respective equilibrium values via the LINCS29 algorithm.

The peptide-bilayer systems simulated in this work were set-up by placing one or two δ-lysin 10-15 Å above the POPC bilayer, followed by solvation with approximately 8000 SPC30 water molecules. Initially, these peptides have parallel orientation with respect to the bilayer surface. A couple of additional simulations were performed with a peptide completely inserted into the bilayer, either vertically or in parallel. Each system was first energy minimized and then went through a 2 ns long position-restraint run in which the peptides are frozen. The production run for each system after the position-restraint simulation is 200 ns long.

 

Results and Discussion

Figure 1 shows the helical wheel diagram of δ-lysin, which clearly expresses the amphipathic nature of peptide with four Lys and three Asp on the hydrophilic face and mostly Leu and Ile on the hydrophobic face. Due to the close proximity, it has been suggested that Lys and Asp form intramolecular salt bridges. δ-lysin also contains four Ser/Thr, which are not typical amino acids found in AMPs or cytolytic peptides. Note that the hydroxyl group at the end of the side chains of Ser and Thr can form hydrogen bonds with the polar lipid head-group. At high concentration, δ-lysin tends to aggregate in solution more than typical AMPs with high helical content (75% helical at 2 μM) due to the absence of net charge. But, at low concentration, oligomers dissociate into dimers and monomers with much lower helicity. A kinetics study by Almeida et al.10 suggested that when δ-lysin binds to a POPC membrane it often forms dimers with very high helical content and eventually translocates the membrane as a trimer. However, it is not entirely clearly whether it is a monomer or dimer in solution that binds to the membrane.

Figure 1.Helical wheel diagram of δ-lysin. Residues with similar hydrophobicity are represented by the same color: hydrophilic or neutral (white), hydrophobic (light gray), aromatic (magenta), negatively charged (red), and positively charged (blue) residues.

To understand the nature of the interaction between δ-lysin and neutral (zwitterionic) lipid membrane, we ran a series of MD simulations (total eight) starting with one or two δ-lysin placed 10-15 Å above the membrane. In three out of four simulations with a δ-lysin monomer, the peptide is oriented in such a way that the hydrophobic face points toward the bilayer. In reality, the monomer in solution is expected to be unstructured or has low helical content under the typical experimental condition. But, due to the limited time scale in our simulations, unfolded peptide will not form a helix on the membrane surface during 200 ns long simulations. Thus, all simulations have a perfectly helical peptide at the beginning. The initial peptide-membrane distance mentioned above provides enough space for the peptide to reorient itself, but it is small enough to maintain the helicity for the most part while the peptide approaches the membrane. For two-peptide simulations, either two monomers or a dimer was placed roughly 10‒15 Å above the bilayer. In addition to the simulations with the peptides placed in the water phase initially, a couple of simulations with a fully inserted helical δ-lysin monomer were performed as well. An overview of all simulations performed in the present work is provided in Table 1.

Table 1.aFor simulations S1-S8, the peptides are initially placed above the bilayer surface. Orientation of hydrophilic face of each peptide is denoted as U (up, facing away from the surface), D (down, facing toward the surface), S (facing side way, parallel to the surface). For simulation S9 and S10, peptide is fully inserted and placed at the core of bilayer either vertically (V) or in parallel (P) with respect to the bilayer surface. bOrientation of peptides when they make full contact with the headgroup (before insertion). cThe second peptide changes its orientation and becomes U once it inserts below the lipid head-group.

POPC and δ-Lysin Monomer Association. For simulations with a single δ-lysin placed in the water phase, all four simulations start with a fully helical δ-lysin with the axis of helix parallel to the membrane surface. Three of those monomer simulations (S2-S4) have the hydrophobic face of δ-lysin toward the membrane surface initially. This choice of orientation is based on the results of our previous work31 on Tp10, a synthetic cell-penetrating peptide, for which the hydrophobic face of peptide preferentially binds to the membrane. In S1, the peptide is initially placed in such a way that hydrophilic/hydrophobic faces point sideway, parallel to the membrane surface.

Figure 2 shows the z-coordinates (i.e. the direction of membrane normal) of δ-lysin from the center of mass of the lipid bilayer as a function of simulation time obtained from simulations S1-S4. In general, peptides approach the membrane slowly and it takes more than 100 ns to pass the lipid head-group region (phosphorous atoms in Fig. 2) in two simulations. In all simulations (S1-S4), δ-lysin approaches the bilayer surface at an angle, with either the C-terminus (S3) or the N-terminus (S1, S2, S4) making the initial contact with the bilayer surface. However, all peptides become parallel to the bilayer surface by the time they are fully engaged in interaction with the lipid head-group. In Figure 3 are shown the snapshots taken from S2 and S3, including the starting peptide-POPC conformation, early stage of association of δ-lysin and lipid head-group, and the final conformation after 200 ns long simulation. The simulation S2 and S3 are chosen for Figure 3 since the behavior of δ-lysin in these two simulations is distinctively different. The peptide in S2 does not penetrate into the membrane at all and is hovering above the bilayer head-group for the entire duration of the simulation. In contrast, the peptide in S3 quickly inserts into the bilayer and penetrates deeply into the core of lipid bilayer.

Figure 2.z-coordinates of the center of mass of δ-lysin monomer (green), the phosphorous atoms (black), and the carbonyl oxygen (red) as a function of simulation time obtained from simulations (a) S1, (b) S2, (c) S3, and (d) S4. The center of mass of lipid bilayer is considered to be z=0.z-coordinates of the center of mass of δ-lysin monomer (green), the phosphorous atoms (black), and the carbonyl oxygen (red) as a function of simulation time obtained from simulations (a) S1, (b) S2, (c) S3, and (d) S4. The center of mass of lipid bilayer is considered to be z=0.

Figure 3.Snapshots taken from simulation (a) S2 and (b) S3. From left to right are shown the initial conformation of the system, first full contact with the bilayer in parallel orientation, and the final conformation after 200 ns. Spheres represent the phosphorous atoms and the gray lines are for the rest of POPC bilayer. For clarity, gray lines are removed in the center and right panels. The side chains of charged residues are also shown in cyan (Lys) and red (Asp).

One of the most surprising observations made for δ-lysin during the early stage of peptide-membrane interaction is that in most cases (S1, S2 and S4) the peptides change their orientation and have the hydrophilic face oriented toward the bilayer surface while they approach the membrane. The only simulation where the hydrophobic face remains oriented toward the bilayer surface is S3. Change in orientation begins after the initial contact with either the C- or N-terminus, and the orientation is adjusted as the other end of peptide approaches the bilayer. The difference in orientation between the peptide in S2 and S3 after binding can be seen in Figure 3 by noting the direction to which the charged side chains point. As the peptides approach the membrane, they start losing helicity in many cases and the amphipathic nature of peptides somewhat diminishes, which makes the distinction of hydrophilic and hydrophobic face more ambiguous. However, peptide orientation can be identified, at least for the earlier part of the simulations, by monitoring the z-coordinates of a few representative side chains, such as Lys and Asp for the hydrophilic face and Ile and Val for the hydrophobic face. As an example, Figure 4 shows the z-coordinates of a few representative side chains located at the middle section of the peptide for S2 and S3 simulations. Specifically, the center of mass of side chains for Asp11, Lys14, Lys22, Ile9, Ile16 and Val20 are plotted as a function of simulation time, where the first three residues are considered as the representative hydrophilic residues and the last three as the hydrophobic residues. It is clear from the figure that the orientations of two peptides are opposite. In S2, hydrophilic residues approach the bilayer surface before the hydrophobic residues (Fig. 4(a)) even though the peptide is initially oriented with the hydrophobic face toward the bilayer surface, which indicates that the peptide turns around in the water phase. The hydrophilic residues are settled in near the lipid head-group, whereas the hydrophobic residues are found above the head-group, mostly exposed in the water phase. Similar peptide orientations are observed in S1 and S4 as well and, as seen in Figure 3, the insertion depths of these peptides are rather shallow. In case of S3, the peptide orientation is maintained and the hydrophobic face of peptide inserts first at around 40 ns.

Figure 4.z-coordinates of the center of mass of the side chains for a few representative hydrophilic and hydrophobic residues obtained from simulations (a) S2 and (b) S3, overlapped with that of phosphorous atoms (black) and carbonyl oxygen (red). For the hydrophilic residues, Aps11 (green), Lys14 (blue) and Lys22 (dark green) are plotted whereas Ile9 (orange), Ile16 (pink) and Val20 (purple) are chosen for the hydrophobic residues.

These observations on the orientation of helical δ-lysin suggest that δ-lysin prefers to have its hydrophilic face to interact first with the lipid bilayer. The preferential orientation of δ-lysin toward the membrane, at least during the initial peptide-membrane association, is clearly different from that of a cell-penetrating peptide Tp10, which was studied by us recently.31 Tp10 has a net +5 charge and consists of mostly Lys and Leu, with no negatively charged amino acids. When Tp10 binds to a POPC bilayer, its hydrophobic face is preferentially oriented toward the bilayer core. It was envisioned that the positively charged Lys residues prefer to stay in water phase due to the initial repulsion from the positively charged outermost layer (choline) of POPC. Hydrophobic face is then working its way toward the membrane core without significant energetic penalty.

The initial attraction between δ-lysin and POPC is mainly driven by the electrostatic interaction, as illustrated in Figure 5, showing the Coulomb interaction and Lennard-Jones (LJ) interaction between δ-lysin and POPC at two different time windows for S2 and S3. During the initial peptide-POPC association (Fig. 5, left panels), the magnitudes of LJ interaction are similar in S2 and S3, but the Coulomb interaction in S2 is significantly larger than that of S3. Thus, it seems that orienting the hydrophilic face of peptide toward the head-group is energetically more favorable when the peptide is still mostly in the water phase, which explains the observed preferential orientation of peptide. δ-Lysin has equal numbers of positively and negatively charged amino acid side chains. The electrostatic repulsion between Lys and the choline group, which causes an energetic penalty in Tp10 case, may be reduced significantly since nearby choline groups will be attracted toward the negatively charged Asp. However, the peptide in S2 does not penetrate into the bilayer interior at all (Fig. 2) and maintain the same orientation throughout the simulation, which is also reflected in the fact that the magnitudes of Coulomb and LJ interactions do not change significantly (Fig. 5, top) during the simulation. Similar behavior was observed for S1 and S4, although the peptide penetration is slightly better than what is observed in S2. On the other hand, δ-lysin in S3 with the hydrophobic side facing the membrane surface quickly inserts into the membrane within 50 ns (Fig. 2(c) and 4(b)) and eventually positions itself below the carbonyl group of lipid tail. During the insertion process, the strengths of both Coulomb and LJ interactions between the peptide and POPC increase drastically (more negative). The magnitude of Coulomb interaction is leveled off at around 120 ns, but the LJ interaction continues to increase, albeit slowly, as it proceeds to the end of 200 ns simulation (Fig. 5, right).

Figure 5.Interaction energy between the peptide and the bilayer obtained from simulation S2 (top) and S3 (bottom) for two different time windows: 0-20 ns (left) and 150-200 ns (right). The interaction energy is divided into the Coulomb (black) and Lennard- Jones (green) contributions.

During insertion, the orientation of δ-lysin in S3 does not change and the side chains of hydrophilic residues, such as Lys and Asp, point toward the water phase (Fig. 3). As the peptide sinks with the hydrophobic face pointing toward the membrane interior, the charged side chains have to go through the lipid head-group region and become tightly bound to the choline (+) or the phosphate (‒) group, depending on their charge. We analyzed the stability of such interaction by computing the minimum distance between two groups, NH3 + (Lys) --phosphate and COO‒(Asp)-choline, as a function of simulation time and the results are reported in Figure 6. For simulation S3, charged side chains complete the salt bridge formation with the lipid head-group later in time, around 40 ns, when they start inserting themselves into the membrane. This is also the time that the Coulomb interaction starts increasing drastically in S3. For S3, all salt bridges are maintained very stable to the end of simulation, which is indicated by very little fluctuation in the phosphate-NH3 + and the choline-COO‒ distances. The peptide is found deep inside the lipid bilayer at the end, but it can maintain strong salt bridges with the choline and the phosphate groups many angstroms away. This is possible in part because of the snorkeling ability of Lys with a long side chain,32-34 which allows the hydrophobic residues to interact with bilayer core, but, at the same time, hydrophilic Lys residue to stay with the headgroup. Such snorkeling behavior of Lys was observed in Tp10 simulations as well, but, as shown in Figure 6, Asp seems to be just as good as Lys in making stable salt bridge with the choline group for δ-lysin. Although snorkeling allows the peptides to be inserted without a significant energy penalty of positioning charged residues in the hydrophobic lipid interior, peptide insertion does cause significant membrane disruption. As shown in Figure 3(b), due to the strong salt bridge formation with the charged residues, the headgroup is dragged into the membrane interior as the peptide moves deeper into the bilayer core.

Figure 6.Minimum distance between NH3 + of Lys and PO4 ‒ of lipid (left column) and between COO‒ of Asp and N(CH3)3 + of lipid (right column) as a function of simulation time obtained from simulations S1-S4 (from the top to bottom panels).

For other simulations shown in Figure 6, where the hydrophilic face of peptide points toward the bilayer interior when they bind to the head-group region, the salt bridge formation is quicker in general, but it is not as stable as seen in S3. The distance fluctuation is noticeably larger and, in some cases, the salt bridge is broken during simulation. Note that these peptides do not penetrate into the bilayer with the orientation they have at the interface. Thus, amphipathicity and the proper orientation of peptide is the key that allows δ-lysin to be inserted into the hydrophobic bilayer interior without an extensive energetic penalty caused by hydrophilic residues buried inside the membrane. Although four simulations are not statistically adequate to make any definite conclusion, given the fact that S3 is the only simulation that shows significant peptide insertion, our simulations suggest that peptide binding to the interface with the hydrophobic residues pointing toward the bilayer core is a necessary condition for peptide insertion, at least for an isolated δ-lysin. The peptides are attracted to the interface initially through the strong electrostatic interaction with the hydrophilic face, but, to be inserted, they must turn around so that they become energetically more favorable inside the bilayer. In fact, the peptide in S4 shows the sign of peptide rotation at the end, which is accompanied by the sudden decrease in the peptide z-coordinate as shown in Figure 2(d).

Finally, we note in passing that the salt bridge type interaction with the lipid head-group is not limited to the charged residues. As briefly mentioned above, δ-lysin contains three Thr and one Ser residue, which are not typically found in AMPs. We found that the side chains of Ser and Thr (with –OH−group) often establish a very stable interaction with the charged lipid head-group (i.e. phosphate). Minimum distance analyses (not shown) reveal that there is one or two Ser/Thr residues (Ser7 or Thr19, in particular) to form exceptionally stable hydrogen bonding with the charged lipid head-group, showing the minimum distance plots that resemble Figure 6. Since Ser7 and Thr19 are located on the hydrophilic face of peptide, such interactions should contribute to the tendency of δ-lysin that preferentially binds to the POPC bilayer with its hydrophilic face.

POPC and δ-Lysin Dimer Association. Unlike most AMPs and cytolytic peptides, δ-lysin has no net charge, which facilitates the aggregation in solution at high concentration (≥ 1 μM). Below the 1 μM threshold concentration, which is the typical peptide concentration employed in binding studies, small clusters of peptides up to tetramer, with the dimer being the most dominant species, are found in solution. 10 Thus, it is important to understand the binding and/or dissociation of δ-lysin dimer on the membrane surface. We performed total four simulations (S5-S8, Table 1) with two δ-lysin interacting with a POPC bilayer. In simulation S5, peptides do not form a dimer initially. Instead, two helical monomers are placed 10-15 Å above the membrane, with opposite orientation with respect to the bilayer surface: one with the hydrophobic face toward the bilayer surface and the other with the hydrophilic face toward the bilayer surface. In this simulation, two peptides are allowed to interact freely and they are monitored if they spontaneously form a dimer on the membrane surface. In S6-S8, a head-to-toe style helical dimer is placed above the lipid head-group region at the beginning of each simulation. In Figure 7 (left column) are shown the initial structures of dimer in S6-S8 (top-down view). In solution, a dimer is expected to be stabilized by the interaction between the hydrophobic faces of amphipathic δ- lysin and the hydrophilic faces are pointing outward and solvated by water.10 Thus, one of the simulations (S8) has an initial dimer with two hydrophobic faces pointing toward each other (Fig. 7(c)). However, such dimer conformation is not expected to be energetically advantageous to reach the bilayer interior. Therefore, in S6 and S7, we tried a dimer with opposite orientation where two hydrophilic faces of δ- lysin pointing toward each other (Fig. 7(a))and (b). Hypothetically, with this peptide orientation, intermolecular salt bridge formation between Lys and Asp is possible. In addition, the Trp residues are in close contact, creating a π- stacking interaction, which can lead to further stabilization of dimer. In S6, the stacked Trp side chains are oriented toward the water phase, whereas they point toward the interface in S7. In each dimer simulation, the initial distances from two monomer units to the bilayer surface are roughly the same and the hydrophilic and hydrophobic faces are pointing sideway along the interface. Since a dimer is less flexible than a monomer, peptides do not tumble within the limited space between the head-group and peptides and, therefore, maintain their orientations during the earlier part of the simulations.

Figure 7.Top-down view of δ-lysin dimer in simulations (a) S6, (b) S7, and (c) S8 at time t = 0 (left) and at t = 200 ns (right). The side chains of important residues are also shown in cyan (Lys), red (Asp) and white (Trp).

Figure 8 shows the z-coordinate of each δ-lysin from the center of mass of the lipid bilayer as a function of simulation time obtained from S5-S8. The peptides in S5, which are monomers at the beginning, quickly insert and penetrate deeper into the membrane than the dimers in other simulations, which implies that it is harder for peptides to go through the head-group region as a dimer. One of the peptides in S5 with the hydrophobic face pointing toward the bilayer interior (labeled as P1 hereafter) inserts deeper than the other monomer with the opposite initial orientation (P2) and positions itself below the carbonyl group of lipid tail within 100 ns. To illustrate the time evolution of two peptides in S5, a few snapshots from S5 are shown in Figure 9. In simulation S5, two monomers attract each other soon after the simulation begins and make contact with the bilayer without changing their orientations. The peptide P1 approaches the lipid head-group in a slanted orientation and immediately inserts into the bilayer (Fig. 9(b)). The peptide P2, on the other hand, stays on top of the head-group region at least for 50 ns. However, P2 eventually follows P1 and inserts into the membrane, albeit slowly. The insertion depths of two peptides do not change significantly after 135 ns. While P2 sinks toward the bilayer interior, it uses the salt bridge with the head-group as a pivot and turns around its orientation so that the charged side chains stay around the head-group region (see the black arrows in Fig. 9(c)). At the end of 200 ns simulation, the charged side chains of P2 point toward the head-group region in general while the hydrophobic residues are buried deeper inside the bilayer. This is what we anticipated from the monomer simulations discussed in the previous section. For P1, there is no need to change its orientation as its hydrophobic face points toward the bilayer interior from the start. Throughout the 200 ns simulation, two monomers are in close contact with each other (C-terminus, in particular). However, intermolecular salt bridge was not found. Instead, most charged residues have salt bridges with the head-group of lipid bilayer. It is also apparent from Figure 9 that the bilayer structure is perturbed extensively by the insertion of two peptides. Not only the head-group of upper leaflet around the peptide is dragged into the core of bilayer, but also the integrity of bottom leaflet is often affected by the peptide insertion during the simulation, which leads to a significant degree of local thinning of bilayer (see Fig. 9(c), red arrows).

Figure 8.z-coordinates of the center of mass of two δ-lysin (blue for P1 and green for P2), the phosphorous atoms (black), and the carbonyl oxygen (red) as a function of simulation time obtained from dimer simulations (a) S5, (b) S6, (c) S7, and (d) S8. The center of mass of lipid bilayer is considered to be z = 0.

Figure 9.Snapshots taken from simulation S5 at (a) t = 0 ns, (b) t = 11 ns, (c) t = 118 ns, (d) t =136 ns. Spheres represent the phosphorous atoms. The side chains of charged residues are also shown in cyan (Lys) and red (Asp). Black arrows in panel (c) indicate the locations of charged side chains of P2. Localized membrane thinning is visible in panel (c) (see double-headed red arrows). Insertion depths of two peptides do not change significantly after 136 ns (panel (d)).

The behavior of δ-lysin dimers in S6-S8 is noticeably different from that of two monomers in S5. Most importantly, the insertion depths are much shallower for the dimers. In fact, the dimers in S6 and S8 do not penetrate into the bilayer at all and they are hovering above the head-group region throughout the entire 200 ns long simulations. The dimers in these two simulations have opposite orientations, one with the hydrophilic faces pointing each other (S6) and the other with the hydrophobic faces (S8), but both dimers are intact for the most part at the end of simulations. The dimer in S6 has an ability to form intermolecular salt bridges, but only very transient salt bridges were found. It seems that the side chains of charged residues tend to establish stable salt bridges with the lipid head-group rather than intermolecular salt bridges. However, the π-stacking interaction between Trp residues persist throughout the simulation. Note that π- stacked Trp residues in S6 are initially in the water phase. For the dimer in S8, only the hydrophobic interactions are possible between two monomer units. Nonetheless, the dimer turns out to be quite stable without visible disruption of head-to-toe style parallel alignment.

Unlike two dimers mentioned above, the dimer in S7 makes significant penetration into the lipid bilayer. However, the dimer falls apart as the simulation begins and, as shown in Figure 8(c), the peptides insert one at a time, not simultaneously. The time evolutions of the z-coordinates of side chains (not shown) revealed that both peptides change their orientations so that most hydrophobic residues buried deeper inside the bilayer (below the carbonyl group). Then, what makes the dimer in S7 behaves differently than that in S6 even though they both have their hydrophilic sides facing each other at the beginning? It is not entirely clear because of the limited numbers of simulations, but one of the main differences between the two dimers in S7 and S6 is the orientation of Trp side chain. In S7, the Trp side chains point toward the bilayer surface, whereas, in S6, they are facing the water phase initially. It has been argued that the hydrophobic residues with aromatic side chains such as Trp and Phe play a key role in membrane penetration by AMPs.35 In fact, mutation of Trp in some AMPs leads to decrease in antimicrobial activity.36,37 In simulation S7, the initial conformation contains the π-stacking interaction between two Trp side chains, but it is not maintained as the integrity of dimer deteriorates. Instead of making π-stacking interaction, each Trp residue is attracted to the interfacial region and buried inside the lipid head-group. The strong preference of Trp residue for the membrane interface is well documented38,39 and it is often believed to be due to the amphipathic nature of Trp side chain.

The results from two peptides simulations allow us to draw a couple of conclusions on the association of δ-lysin and POPC bilayer. First, the inspection of Figure 7 and Figure 8 suggests that peptide insertion as a dimer is less favorable than as a monomer, regardless of the dimer orientation. Under the typical experimental condition, δ-lysin exists in solution as a monomer or a small oligomer (up to tetramer). The dimer in solution is most likely stabilized by the hydrophobic interaction similar to the dimer in S9. However, such dimer does not penetrate into the bilayer as seen in Figure 8(d). Based on the binding kinetics experiment of δ-lysin, Almeida and coworkers10,23 suggested that the dominant membrane bound species is dimer, which is most likely stabilized by electrostatic interaction and Trp π- stacking interaction. The dimers in S7 and S8 resemble the structure proposed by Almeida et al., but, again, they do not penetrate deep into the membrane as a dimer. The results from our MD simulations, in conjunction with the experimental observation, suggest that the binding of δ-lysin on POPC occurs in two steps: peptide insertion as a monomer followed by dimerization. A dimer in solution may be associated with the bilayer, but it must dissociate into monomers before it inserts into the bilayer interior. It has been suggested from free energy calculations40 as well as our previous MD study31 of Tp10 that the preferential location of membrane bound amphipathic peptide is often found between the lipid tail and the head-group region with the hydrophobic face of peptide pointing toward the bilayer center. Therefore, once δ-lysin monomer is settled in below the headgroup region, subsequent self-association of peptide may occur as the concentration of membrane bound monomer increases.

Another important observation made in two-peptide simulations is that the formation of intermolecular and intramolecular salt bridges is far less likely than what was assumed in experimental studies. It has been suggested that the structure of δ-lysin should allow the head-to-toe (or antiparallel) style dimer to have intermolecular salt bridges, Asp4-Lys26 and Asp11-Lys22 in particular. Moreover, formation of intramolecular salt bridges has been hypothesized to explain the membrane binding affinity of δ-lysin that is much higher than what is predicted by the Wimley-White interfacial scale.24,25 However, our MD simulations show that both intermolecular and intramolecular salt bridges are rare. Inspection of MD trajectories shows that the motions of side chains of charged residues, especially Lys, are very floppy and the formation of salt bridges between charged residues are transient at best. As a result, stable intra- and intermolecular salt bridges do not last, if any, for more than a few nanoseconds.

POPC with Fully Inserted δ-Lysin. As seen in the simulations discussed in the previous sections, δ-lysin monomer and dimer initially placed in the water phase does not penetrate into the bilayer core within the time scale of our simulation. Only the monomer with its hydrophobic face pointing toward the membrane interior shows a significant insertion depth. But even such peptides do not reach the center of the bilayer and tend to be settled in between the carbonyl group and the lipid tail. This is because the salt bridges formed between the charged residues and the headgroup are strong enough to prevent a peptide from further insertion. In order for a peptide to translocate across the bilayer, as suggested in recent experiment, these salt bridges must be broken, which is believed to be a stochastic process. Increasing peptide concentration would facilitate the peptide insertion due to the increased mass imbalance between two layers of lipid. In fact, Marrink and coworkers41 showed through MD simulations that magainin-2 forms a disordered toroidal pore when the peptide density is around P/L=1/32. However, the peptide densities that induce pore formation in magainin-2 simulations are much higher than what is normally adopted in experiment. In general, at lower concentration that is more relevant in experiment, neither a spontaneous pore formation nor a complete insertion of peptide is observed in MD simulations.

In the present work, we performed a couple of simulations with a single δ-lysin already inserted into the bilayer core in order to understand the structure and stability of a peptide inside the membrane. Two different conformations of δ-lysin were investigated: vertical and parallel insertion. We adopted the same simulation protocols as in simulations S1- S8 and the initial peptide conformation is fully helical. Figure 10 shows the snapshots taken at t = 0 (left column) and t = 200 ns (right column) from simulation S9 (top, parallel insertion) and S10 (bottom, vertical insertion). The panels on the left are the final conformations after 2 ns position restraint simulations.

When a helical δ-lysin is inserted into the bilayer center in parallel orientation, the intrusion of water and lipid headgroup toward the peptide is extensive after the position restraint run. This leads to significant disorganization of bilayer, but, it is necessary to prevent the charged residues from being exposed to hydrophobic lipid core, which is highly unfavorable energetically. Notice that the hydrophilic face of peptide is pointing toward the upper leaflet and only the upper leaflet experiences noticeable disturbance. This configuration allows the charged side chains, Lys and Asp, to interact with the head-group even though they are buried deep inside the membrane. Once the simulation begins the peptide quickly moves toward the upper leaflet and, at the same time, the integrity of bilayer is restored for the most part. The water molecules found inside the bilayer at the beginning are mostly retracted to the water phase outside the bilayer. The final location of peptide is between the headgroup and the lipid tail, which is consistent with the simulations where the peptide is inserted from the water phase (see Fig. 3(b) and Fig. 9(d)). Thus, this result implies that the final configuration of membrane bound δ-lysin monomer should resemble the state described in Figure 10(a) (right). It should also be noticed that the helicity of peptide in S9 is maintained almost perfectly even after 200 ns simulation. In fact, the helicity of membrane bound δ-lysin measured by circular dichroism is close to 100%,24 which is consistent with the result of our MD simulation. As shown in Figure 10 (a), the helix is slanted with respect to the bilayer surface with the C-terminus inserted deeper. This is due to the fact Lys residues are clustered at the C-terminus. The long Lys side chains can do snorkeling and, therefore, the C-terminus can be buried deeper inside the membrane. It is interesting to see that not only the phosphate groups in the upper leaflet (because of the salt bridges with the charged residues), but also those in the lower leaflet are dragged toward the peptide. This causes significant localized thinning of lipid bilayer (see Fig. 10(a) and Fig. 9(c)).

In case of vertically inserted helical δ-lysin, there exists noticeable decrease in helicity of peptide after 200 ns simulation. The C-terminus side with three Lys is in close contact with the head-group of upper leaflet, whereas the Lys in the middle and three Asp residues are dragging the head-group into the bilayer interior. The water molecules initially found in the hydrophobic core are mostly pushed out of the bilayer. The final conformation of peptide is significantly tilted and stretched (due to the loss of helicity) because of the mismatch between the length of δ-lysin and the thickness of bilayer. This is in contrast with the behavior of vertically inserted shorter peptide, 21-residue Tp10, which maintains its helicity as well as the vertical orientation.31 Although the spontaneous translocation of δ-lysin as monomer or dimer was not observed in our MD simulation, two simulations with fully inserted δ-lysin monomer (Fig. 10) allows us to speculate a possible mechanism of monomer translocation, if it occurs. At first glance, the final structures of peptides from two simulations, shown in the right column of Figure 10(a) and (b), look very different. However, a transition from the membrane bound peptide (Fig. 10(a)) to the vertically inserted peptide (Fig. 10(b)) can occur if the salt bridges between the Lys residues at the C-terminus (inserted deeper than the N-terminus) and the head-group are broken and the Lys side chains are simply flipped over to point the headgroup of bottom leaflet, which is already dragged toward the peptide. Breaking a stable salt bridge is a stochastic process and energetically unfavorable. However, as the peptide concentration increases, mass imbalance between the outer and inner layer of membrane should increase the degree of membrane disruption and improve the chance of breaking salt bridges.

Figure 10.Snapshots taken from simulations (a) S9 and (b) S10. δ- lysin monomer is fully inserted either vertically (S9) or in parallel (S10) with respect to the membrane surface. The initial conformation is shown on the left and the final conformation on the right. Larger spheres (gold) represent the phosphorous atoms. To illustrate the degree of water penetration, oxygen and hydrogen atoms of water molecules are represented by small red spheres and white spheres, respectively. The side chains of important residues are also shown in cyan (Lys), red (Asp) and white (Trp).

 

Conclusion

We performed a series of molecular dynamics (MD) simulations of hemolytic peptide δ-lysin monomer and dimer interacting with a zwitterionic POPC bilayer. Total eight simulations were carried out with δ-lysin monomer or dimer placed in the water phase above the bilayer head-group. Although the number of simulations is too small to draw a definite conclusion, our simulations showed that the peptide insertion into the bilayer occurs as a monomer. No insertion as a dimer was observed. In addition, stable intra- and intermolecular salt bridges between charged residues, which was suggested in recent kinetic studies, was not found. For a dimer to be inserted into the membrane, the dimer conformation needs to be broken and the subsequent insertion occurs as a monomer. When a helical monomer binds to the interface, the insertion depth depends on the orientation of amphipathic δ-lysin. Specifically, the hydrophilic face of peptide is preferentially attracted to the lipid head-group once the simulations begin, primarily due to the large contribution from the electrostatic interaction. This is in contrast with the behavior of Tp10, a Lys rich cell-penetrating peptide, which we studied recently.31 Note that Tp10 has no negatively charged residue and net +5 charge, whereas δ- lysin has equal numbers of positively and negatively charged residues.

The monomer with its hydrophilic face interacting first with the lipid head-group, however, does not penetrate into the bilayer. Only those monomers with the hydrophobic face oriented toward the bilayer core at the interface showed significant insertion depths, primarily due to the favorable hydrophobic interaction with the lipid tail. The inserted δ- lysin monomer ends up between the lipid head-group and the tail, most likely around the carbonyl group. We speculate that the membrane bound dimer, which was assumed in kinetic analysis by Almeida and coworkers,4,23 can form once the concentration of monomer around the carbonyl group reaches a certain threshold. During insertion, the side chains of charged residues form stable salt bridges with the choline or the phosphate of lipid head-group. This makes it possible for the charged residues to be buried inside the membrane, by dragging the head-group into the hydrophobic interior, as the monomer sinks toward the center of bilayer. In particular, the snorkeling of long and flexible side chains of Lys residues, which are clustered at the C-terminus of δ-lysin, allows the C-terminus to go deeper into the membrane, resulting in a tilted orientation with respect to the bilayer surface (Fig. 10(a)). Although a complete insertion of peptide was not observed in 200 ns simulation, two simulations with a pre-inserted monomer suggest a mechanism for the membrane bound monomer (Fig. 10(a)) to become a fully inserted state (Fig. 10(b)). It seems that a complete insertion of monomer requires breaking of stable salt bridges between Lys and the phosphate group, which is more likely to occur stochastically with higher mass imbalance between two leaflets of bilayer. Then, the free Lys side chains can flip and become connected to the head-group of bottom leaflet, which are already pulled noticeably toward the membrane core (Fig. 10(b)).