Effects of Peptide Charge, Orientation, and Concentration on Melittin Transmembrane Pores

Abstract

Melittin is a short cationic peptide that exerts cytolytic effects on bacterial and eukaryotic cells. Experiments suggest that in zwitterionic membranes, melittin forms transmembrane toroidal pores supported by four to eight peptides. A recently constructed melittin variant with a reduced cationic charge, MelP5, is active at 10-fold lower concentrations. In previous work, we performed molecular dynamics simulations on the microsecond timescale to examine the supramolecular pore structure of a melittin tetramer in zwitterionic and partially anionic membranes. We now extend that study to include the effects of peptide charge, initial orientation, and number of monomers on the pore formation and stabilization processes. Our results show that parallel transmembrane orientations of melittin and MelP5 are more consistent with experimental data. Whereas a MelP5 parallel hexamer forms a large stable pore during the 5-μs simulation time, a melittin hexamer and an octamer are not fully stable, with several monomers dissociating during the simulation time. Interaction-energy analysis shows that this difference in behavior between melittin and MelP5 is not due to stronger electrostatic repulsion between neighboring melittin peptides but to peptide-lipid interactions that disfavor the isolated MelP5 transmembrane monomer. The ability of melittin monomers to diffuse freely in the 1,2-dimyristoyl-SN-glycero-3-phosphocholine membrane leads to dynamic pores with varying molecularity.

Introduction

Membrane-active peptides (MAPs) interact with and disrupt cell membranes. They include natural toxins and antimicrobial peptides, which are produced widely in nature as an innate mechanism of defense against bacterial and viral infections (1, 2, 3). Over 1000 MAPs have been identified to date, and there is a growing interest in their potential pharmacological applications as antibiotics, antivirals, and anti-inflammatory agents (4, 5). Multiple experimental studies have shown that MAPs exert their cytolytic activity by their action on the cell membrane, including pore formation, alteration of the osmotic balance, and membrane disintegration (6, 7, 8). Membrane binding and disruption occurs without the mediation of protein receptors, which greatly hinders the development of resistance against these peptides. MAPs are predominantly cationic and thus bind selectively to anionic bacterial membranes. Although the experimental data on these peptides are extensive, the structural details of the peptide-membrane interaction are still unknown. This information is essential for gaining a complete understanding of their mechanism of action and extending their use toward clinical applications.

The 26-residue peptide melittin is the main component of bee venom. It is one of the most widely studied helical MAPs because of its potent antimicrobial, antiviral, anticancer, and hemolytic activities (9, 10, 11, 12, 13). Melittin is mainly unstructured in solution but forms a helical tetramer under certain conditions of peptide concentration, pH, temperature, and ionic strength (14, 15, 16, 17). It binds and inserts into lipid bilayers as an amphipathic helix (9, 18, 19). Dye leakage, fluorescence energy transfer, and neutron in-plane scattering experiments, among others, have revealed that melittin forms transmembrane pores beyond a threshold concentration (peptide/lipid (P/L^∗)) on the bilayer surface (20, 21, 22, 23). However, the detailed structure of a melittin transmembrane pore at atomic resolution has not yet been solved. Experimental evidence indicates that melittin forms pores of the toroidal type, in which both the peptides and hydrophilic lipid headgroups line the transmembrane water channel with sizes ranging from 3 to 4.5 nm and structural assemblies of four to seven peptides (21, 24, 25, 26). Transmission electron micrographs showed melittin pores as ring-like structures in zwitterionic vesicles, whose sizes and further cross-linking analysis agreed with the proposed molecularity of four to eight peptides per pore (26). Quasielastic neutron scattering experiments have recently shown that melittin promotes important changes in lipid dynamics and phase behavior at concentrations below P/L^∗, which highlights melittin’s potent disruptive effects in the host membrane beyond the generation of transmembrane pores (27).

Computational methods have reached beyond the resolution of experimental approaches to describe not only structural details at atomic resolution but also provide information on the specific intermolecular interactions that generate and support these pores. Molecular dynamics (MD) simulations have been used extensively to investigate the pore-formation process by melittin. Different works have provided valuable insights into the structure and stability of the toroidal pores (28); the preferred orientation of the peptides inside the bilayer (29); and the effects of peptide concentration and helicity on membrane binding, insertion, and translocation (30), among others. These studies have also examined in detail the importance of the charged groups in membrane binding and pore formation. In a series of simulations, our group further determined the influence of the distribution of charges in the melittin sequence on the formation of toroidal pores (31). In this work, K7 in the N-terminal region of melittin was shown to be essential for the formation of toroidal pores independently of the positively charged residues in the C-termini. The substitution of some of these C-terminal charges for amino acids that enhance melittin amphipathicity also resulted in cylindrical instead of toroidal pores. More recently, our group examined the pore formation of a melittin parallel tetramer in zwitterionic (1,2-dimyristoyl-SN-glycero-3-phosphocholine (DMPC)) and partially anionic (DMPG/1,2-dimyristoyl-SN-glycero-3-phosphoglycerol 3:1) membranes by multimicrosecond-long MD simulations (32). The results in DMPC described at an atomic resolution the formation and stabilization of a dynamic toroidal pore in the lower end of the experimentally determined size range, supported by four melittin monomers.

In an effort to improve melittin’s pore forming and cytolytic capabilities, Wimley and collaborators performed a high-throughput screen of more than 7000 peptides and found a few potent gain-of-function variants (33). Among these, the synthetic peptide MelP5 dramatically enhanced leakage at P/L ratios as low as 1:1000, 10-fold more potent that wild-type melittin. MelP5′s increased activity comes from the replacement of five amino acids, which improves helicity and amphipathicity. Three of these changes eliminate positive amino acids in the C-terminal region (R22A, K23A, R24Q), reducing MelP5’s net charge to +2 at neutral pH. This feature and the overall sequence similarity with melittin makes MelP5 an excellent candidate for exploring the effect of the peptide charge on pore formation and stabilization by these peptides. MelP5 forms pores in neutral phosphorylcholine (PC), in mixed PC/phosphorylglycerol, and PC/cholesterol membranes, but the structure of these pores has not yet been determined (34). Single-ion-channel recordings have shown that MelP5 forms both transient and stable pores, with an estimated molecularity of 10–12 peptides per pore, whereas melittin forms only transient pores with a molecularity of 3–9 (35). A recent computational study based on both coarse-grained and all-atom MD simulations has examined the differences in aggregation, insertion, and pore structure between MelP5 and melittin (36). They found that only MelP5 was able to aggregate and insert into the lipid bilayer at low peptide concentration. The authors concluded that the reduced charge of MelP5′s C-terminal region decreases electrostatic repulsion and favors pore formation by increasing the hydrophobic interactions between monomers. Earlier studies, however, showed that truncation of all or part of the C-terminal region results in decrease in activity (37, 38). Apparently, truncation and neutralization produce different effects.

In this work, we have performed all-atom MD simulations on the microsecond timescale of independent hexamers of melittin and its gain-of-function variant MelP5 to determine the effect of the peptide charge on pore formation and stabilization. In addition, we have simulated a melittin octamer to examine the influence of initial oligomeric state and individual MelP5 peptides in transmembrane orientation. Motivated by previous results on magainin family peptides (39), we also consider antiparallel peptide orientations for the hexamers. The results provide insights into the sequence determinants of pore-forming peptides and the origin of the enhanced activity of the MelP5 variant.

Methods

The structure of melittin was obtained from the Protein Data Bank (PDB: 2MLT), and MelP5 was built as an ideal α-helix. The peptides in the three systems were set up with the axis parallel to the membrane normal in a tight bundle using the CHARMM program (40). All systems were simulated at a constant 1:30 P/L ratio and in an either parallel or antiparallel peptide orientation. The membrane-peptide systems were built and solvated using the CHARMM-GUI server (41). All the systems were inserted on pure DMPC membranes. In the case of the melittin octamer, the empty space inside the helical bundle was filled with water molecules. The “MelP5 to melittin” run corresponds to the last frame of the MelP5 parallel-hexamer simulation with the peptides mutated to melittin (A10T, A22R, A23K, Q24R, and L26Q) using the program Chimera (42), followed by a regular equilibration process. The MelP5 dry trimer was built from the structure of the melittin dry trimer in the last frame of the melittin parallel-hexamer simulation. Again, the structure was mutated using Chimera and was further equilibrated in NAMD before the production run.

The initial energy minimization and equilibration processes were performed locally with the program NAMD (43) using the CHARMM c36 force field (44). Production simulations were run on Anton 1 and Anton 2 Supercomputers (Pittsburgh Supercomputing Center, Pittsburgh, PA) following the protocol used in our previous work with the following minor alterations (32). The set-up information for every simulated system is provided in Table 1. Nonbonded-interaction-cutoff values were determined automatically by Anton; the temperature ranged between 300 and 310 K. The conversion of the trajectory files and their visualization as well as the calculation of the solvent-accessible surface area were done with the VMD program (45). The analysis of the simulations was carried out with the programs CHARMM (interaction-energy calculation, pore-radius size, and tilt-angle values) and CPPTRAJ (helicity, hydrogen bonds) (46). We have considered that a lipid headgroup or water molecule is within the pore if the phosphorus and O2 atom, respectively, are within 10 Å of the membrane midplane. Pore-radius values correspond to the water-channel radius in each case.

Table 1.

Simulated System Characteristics

Peptide	Initial Orientation	Number of Peptides	Number of Lipids	Simulation Time (μs)	System Size (Å)
Melittin	Parallel	6	180	5	88 × 88 × 79
Melittin	Parallel	8	240	9	102 × 102 × 69
MelP5	Parallel	6	180	5	88 × 88 × 77
MelP5 to melittin	Parallel	6	180	5	88 × 88 × 77
Melittin to MelP5	Parallel	3	90	5	58 × 58 × 83
MelP5	N/A	1	90	0.2–1	58 × 58 × 83
Melittin Q26L	Parallel	6	180	5	88 × 88 × 77
Melittin	Antiparallel	6	180	5	88 × 88 × 79
MelP5	Antiparallel	6	180	5	88 × 88 × 77

All systems start from a tight bundle except the “MelP5 to melittin” trimer and single-peptide simulations. N/A, not applicable.

Results

In this work, we have simulated different oligomeric states of melittin and MelP5 to examine the effects of peptide net charge and molecularity on pore formation and dynamics. We also examine the effect of the orientation of the monomers in the initial helical bundle (parallel versus antiparallel). Membrane lipid composition and P/L ratio were identical in all simulations. At this P/L ratio (1:30) and in this lipid, both peptides are expected to be fully transmembrane (24). In this section, we provide a description of the most significant events taking place during the trajectories. In all cases, a pore opens up right at the start of the simulations, if it is not there already (octamer). We are following the terminology used in previous studies (39, 47) with regards to peptide orientation; i.e., we refer to the S-state (surface) for tilt angles in the range 60–120°, T-state (tilted) for 30–60°/120–150°, and I-state (inserted) for 0–30°/150–180°. The results of the last five simulations in Table 1 are given in the Supporting Materials and Methods.

The durations of most simulations reported here are from 5 to 9 μs. This timescale is clearly inadequate for sampling the very heterogeneous configurational ensemble of melittin-bilayer systems. It should, however, be sufficient for convergence to a local free-energy minimum. These simulations allow us to examine the viability of various putative pore structures, that is, if a structure breaks down within 5 μs, it cannot belong to the actual configurational ensemble. We then analyze the simulations to obtain insights into the types of interactions that stabilize or destabilize these putative pore structures.

Melittin parallel hexamer

The initial, equilibrated structure has six peptides tightly packed in the I-state around a narrow water pore. After the first 100 ns of the simulation, the pore momentarily closes (Fig. 1 A), and two peptides separate from the rest. The pore reopens shortly after and grows rapidly to ∼10 Å, but only four of the six initial peptides surround the water channel at this time. The remaining two monomers (violet and orange in Fig. 2) are inserted in the bilayer and associate tightly with one of the peptides in the pore (blue in Fig. 2). The formation of this trimeric assembly protects the polar face of these peptides from the hydrophobic surroundings. At around 2 μs, this trimer dissociates from the water channel and diffuses freely within the membrane. It does not lead to the formation of a second pore and remains as a “dry trimer” for the entire simulation time. At 4.7 μs, the dry trimer approaches the pore to restore the tetrameric configuration observed at the beginning of the simulation, but only until 5 μs, when it separates once more from the water channel. During this time, a slight increase in pore size is observed as a consequence of the brief addition of an extra peptide to the water channel (Fig. 1 A). The peptides in the dry trimer remain in an I-state configuration for the whole trajectory, with average tilt-angle values ranging from 9 to 19° during the last microsecond of the trajectory. On the other hand, the monomers in permanent contact with the aqueous pore sample a wider range of tilted orientations, with average tilt-angle values of 20–50° during the last microsecond (Fig. 3 A).

Figure 1.

Evolution of the pore-radius size along the different simulations. (A) Melittin parallel-hexamer. (B) Melittin parallel-octamer. (C) MelP5 parallel-hexamer simulation. (D) MelP5 to melittin parallel-hexamer.

Figure 2.

Progressive separation of the dry trimer in the melittin parallel hexamer. Whereas the monomers in the water pore do not interact, the right side of the figure shows the extensive intermolecular network of interactions formed by the peptides in the dry trimer. Hydrogen-bond occupancies are shown in Table S2. To see this figure in color, go online.

Figure 3.

Tilt-angle values for the different monomers in each trajectory during the simulation. (A) Melittin parallel-hexamer. (B) Melittin parallel-octamer. (C) MelP5 parallel-hexamer. (D) MelP5 to melittin parallel-hexamer. The colors in the plot correspond to the colors in which each monomer is displayed in Figs. 2, 4, 5, and 6. To see this figure in color, go online.

There are no long-lasting peptide-peptide interactions between the monomers supporting the water channel, only the formation of transient hydrogen bonds between the C-terminal Q25/26 residues and K21 (Fig. 2). On the other hand, the dry trimer is stabilized not only by shielding the peptides’ hydrophilic face in a tight bundle but also by multiple interactions between the charged and polar amino acids in their C-terminal region (Figs. 2 and S1). The average interaction energies over the last microsecond of the trajectory between the C-terminal regions (residues K21–Q26) range between −7 and −5 kcal/mol. This assembly results in more favorable peptide-peptide interaction-energy values for those monomers in the dry trimer during the last microsecond of the simulation (−35 ± 8 to −23 ± 6 kcal/mol) than for those around the water channel (−11 ± 10 kcal/mol). There is some minor unfolding of the C-terminal region of some of the monomers, but the average overall helicity during the last microsecond remains above 70%.

Melittin parallel octamer

In the case of the melittin octamer, a water channel was built in the initial structure to fill the empty central space in the helical bundle. Several lipid headgroups insert into the membrane rapidly after equilibration, which leads to a fast increase in pore size. Starting at 40 ns, three monomers dissociate sequentially from the water channel (Fig. 4). Two of these peptides, colored red and gray in Fig. 4, dissociate in an I-state conformation, whereas the third one, colored orange, moves toward the upper leaflet to adopt a surface orientation before leaving the pore. The three peptides maintain their orientation for the remainder of the simulation, with average tilt-angle values of 17° (red monomer), 113° (orange monomer), and 28° (gray monomer) during the last microsecond of the trajectory. The rest of the melittin peptides around the pore mainly adopt an I-state configuration or occasional highly tilted orientations, with average tilt-angle values between 13 and 50° (Fig. 3 B). The transient reorientation events bring together two or more peptides that form interactions at the C-terminal region between the charged side chains of R22/24 and Q26 (Fig. 4). All peptides show some unfolding in either or both the N- and C-termini, where the majority of the charged residues are located. The final average helicity ranges between 66 and 80%. Despite the loss of three of the monomers, the pore radius reaches a maximum value of 14–15 Å during the first microsecond and maintains it for the remainder of the simulation (Fig. 1 B).

Figure 4.

Dissociation of individual peptides from the initial melittin octamer. The intermolecular interactions observed at 9 μs are shown in detail in the right side of the image. In all the figures, water molecules are represented as lines, melittin monomers as cartoons, and lipid headgroups as orange spheres. The rest of the lipid molecules have been omitted for clarity. Hydrogen-bond occupancies are shown in Table S2. To see this figure in color, go online.

MelP5 parallel hexamer

This MelP5 parallel hexamer forms a stable wide pore supported at all times by the six monomers and multiple lipid headgroups (Fig. 5). The pore grows for the first 2 μs to reach a final radius of 14 ± 1 Å (Fig. 1 C). Similar to the behavior observed in the melittin pores, the MelP5 peptides around the water channel explore a wide range of orientations with average tilt angle values between 15 and 52° during the last microsecond of the trajectory (Fig. 3 C). On average, the tilt angles of MelP5 are higher than those of melittin (39 vs. 25° over the last μs). The pairwise interactions along the trajectory are mostly dependent on the separation distance between the monomers. We observe a transient network of hydrogen bonds between N-terminal and C-terminal amino acids that lead to interaction-energy average values in the range of −2.5 ± 3.5 to −9 ± 8 kcal/mol during the last microsecond of the trajectory. The average helicity is over 80%, with the exception of one monomer that unfolds at the C-terminus, dropping to an average value of ∼75% (green in Fig. 5).

Figure 5.

Intermediate and last frame of the MelP5 parallel-hexamer simulation. To see this figure in color, go online.

MelP5 to melittin hexamer

To confirm and gain insight into the different behaviors observed for the melittin and MelP5 parallel hexamers, we continued the MelP5 parallel-hexamer simulation but with the peptides mutated back to melittin. During equilibration and the first steps of the production, one of the monomers (gray in Fig. 6) detaches partially from the water channel. The monomer maintains its contact with the pore mainly through the N- and C-terminal regions until it completely separates from it at ∼40 ns. The dissociated monomer adopts a stable I-state orientation for the remainder of the simulation, whereas the peptides around the pore sample different orientations along the trajectory. However, during the last microsecond, all the monomers but one (blue in Fig. 6) are stabilized in an inserted orientation with average tilt-angle values between 15 and 28° (Fig. 3 D). The rapid separation of one monomer and the formation of the final pentamer resulted in the sharp decrease in pore size observed at the beginning of the simulation (Fig. 1 D); within the first 100 ns, the pore radius decreased from ∼15 Å to less than 10 Å. Afterwards, the pore grew and fluctuated around values close to the starting one, until it stabilized to an average size of 11 ± 2 Å during the last microsecond (Fig. 1 D). The pore structure at the end of the trajectory is stabilized by only a few intermolecular interactions between melittin monomers, composed mainly of the C-terminal Q26, R22, and K23 (Fig. 6). The pore is wide enough to allow the insertion of multiple lipid headgroups that further participate in the stabilization of the monomers around the water channel. Average helicity values during the last microsecond resemble those obtained in other melittin simulations, ranging from 67 to 74% for the peptides in the pore and 80% for the dissociated one.

Figure 6.

Rapid dissociation of one of the melittin monomers (gray) in the “MelP5 to melittin” parallel system. The pore size decreases slightly during the simulation, which favors the contacts shown in the right side of the figure between some of the monomers in the pore. Hydrogen-bond occupancies are shown in Table S2. To see this figure in color, go online.

Discussion

Despite decades of experimental and theoretical work, we are still struggling to understand how MAPs cooperate with lipids to shape membrane pores. In melittin, numerous sequence modifications have been made in hopes of deciphering the determinants of its function, but a clear picture is lacking (10). A key question is the role of charged amino acids. In previous studies, our group has identified residues in melittin that influence pore shape (31, 48) and examined the structure of melittin tetramers in neutral and partially anionic membranes by all-atom MD simulations on the microsecond timescale (32). We have now extended this work on melittin by examining the role of peptide charge, mutual orientation, and molecularity in pore stability and dynamics. To this end, we compare the outcome of microsecond-long simulations of melittin and MelP5 parallel or antiparallel hexamers and a parallel octameric melittin system.

In general, the placement of the peptides as a transmembrane bundle leads to rapid insertion of water molecules into the hydrophobic core of the bilayer. In most of our simulated systems, this process is also accompanied by the insertion of several lipid headgroups into the membrane, in agreement with experimental evidence of melittin forming toroidal pores (21, 30). In all systems simulated (except for the MelP5 antiparallel hexamer), the pore gets larger as the number of monomers around it increases. This finding also agrees with experimental studies that linked increasing peptide concentrations with larger pores (21). Lastly, the overall helicity of the melittin peptides in all the simulations is within a reasonable range of the reported experimental values in zwitterionic membranes, including DMPC (49).

MelP5 is a gain-of-function synthetic peptide obtained by Wimley and co-workers from a high-throughput screening of an extensive library of melittin variants (33, 34). Besides its increased helicity and amphipathicity, the reduced positive charge of MelP5 could play a role in the higher activity of this peptide. Our results show significant differences in pore structure depending on the initial orientation of MelP5 inside the membrane. The parallel hexamer generates a large and stable toroidal pore that fluctuates minimally around 14 Å radius, whereas the antiparallel orientation turns into a tight structure with no involvement of lipid headgroups surrounding a very small water channel (∼6 Å radius) (see Supporting Materials and Methods). Clearly, the parallel orientation forms a pore in better agreement with the experimental data on MelP5 activity. In both MelP5 systems, however, all the peptides are in contact with the water channel during the entire simulation time.

In contrast, a progressive dissociation of one or more peptides from the pore is observed in all our melittin simulations. A detailed look at the trajectories suggests that this is not due to electrostatic repulsion between the peptides. Some of the peptides dissociate from the pore later in the simulation, after losing contact with their neighboring monomers. To further test if the initial proximity of the melittin helices (and therefore the potential electrostatic repulsion) drives the separation of one or more monomers, we mutated the peptides in the last frame of the MelP5 parallel hexamer simulation into melittin and ran a new 5-μs simulation starting from this configuration, in which the six monomers are quite far apart with barely any interactions between them. Although the MelP5 pore structure was stable for the entire initial simulation, the mutated melittin version dissociated into a pentamer plus a single peptide ∼50 ns after the beginning of the new trajectory (Fig. 6). By the end of the simulation, the pore size reached a similar value to that obtained in the melittin parallel- and antiparallel-hexamer simulations (10–11 Å radius). The closer proximity of the peptides during the last microsecond of the trajectory lowers the peptide-peptide interaction energies for the five monomers around the pore from initial values of ∼0 to average values between −8 and −1.5 kcal/mol at the end of the simulation. The interactions observed between these neighboring peptides in the final pore structure involve R22 and Q26 (Fig. 6), just as in the other melittin simulations. These results indicate that electrostatic repulsion between peptides is not the major driving force for the instability of melittin pores. This argument is reinforced by the partial reorientation toward a parallel orientation observed in the melittin antiparallel-hexamer simulation and the interactions of R22 and R24 with the side chains and backbone atoms of the C-terminal Q25 and Q26 (see Supporting Materials and Methods).

Additional evidence for this counterintuitive lack of electrostatic repulsion comes from the formation of what we refer to as a “dry trimer” in the melittin parallel-hexamer simulation. The three peptides come together, shielding their polar faces from the hydrophobic core of the membrane (see Table S3). We have examined the potential role of the small polar amino acids T10, T11, and S18 in the aggregation of these three monomers, as polar amino acids have been shown to drive the association of transmembrane helices (50, 51). In our simulations, however, the side chains of T10, T11, and S18 are engaged in intrahelical hydrogen bonds, and their pairwise interaction energies are negligible. However, there is an extensive hydrogen-bonding network between the charged lysine and arginine residues (K21, R22–24) and Q26 (Figs. 2 and Document S1. Supporting Materials and Methods, Figs. S1–S6, and Tables S1–S5, Document S2. Article plus Supporting Material; Table S2). There are almost no ions or water molecules near this network. The overall interaction energies between the C-termini are favorable, with maximal values of −13.5 ± 5 kcal/mol. To further dissect these interactions, we examined the contributions to the interaction energy of different pairs of charged amino acids. We do not observe H-bonds between pairs of charged peptides, but the average interaction energy values for the different charged amino acids pairs over the last microsecond range from −2.2 to 0.8 kcal/mol (Table S4). The few slightly unfavorable interactions are overcome by the favorable interactions obtained between the whole C-termini, including the polar residues Q25 and Q26 (−13.5 to −2.5 kcal/mol). It appears that the flexible charged side chains manage to find configurations that minimize the repulsion between them while making favorable interactions with polar ones. A similar observation was made in a recent study of oligoarginine peptides (52). The association of these peptides was found to be mediated by stacking of the arginines and formation of salt bridges to the carboxyl termini. Previous computational studies of charged-side-chain interactions also showed an attraction between stacked arginines but not lysines (53, 54, 55, 56). Additional analysis and discussion of this issue is given in the Supporting Materials and Methods, Text S2.

If not electrostatic repulsion between the peptides, what then causes the breakup of the melittin hexamer in contrast to that of MelP5? To investigate this question, we computed average interaction energies with the surrounding waters and lipid molecules for peptides on the edge of a pore and for peptides surrounded by lipids. In the absence of dissociation events in the MelP5 pore trajectories, we ran a 1-μs simulation of a MelP5 monomer starting from an ideal helix (see “MelP5 single peptide” in Supporting Materials and Methods). During this simulation, the peptide first adopted a large tilt (0.2 μs) and then reduced the tilt while the C-terminal region unfolded and moved toward water (1 μs) (Fig. S2). The results in Table 2 show that the transfer of melittin from the pore to the membrane interior is accompanied by loss of interactions with water and gain of favorable interactions with the lipids, as expected. Upon the same transfer, the tilted, more helical MelP5 structure (0.2 μs) experiences much less gain in lipid interactions and an increase in intramolecular energy due to partial loss of helicity; melittin is 67% helical in the pore and 80% in lipids, whereas MelP5 is 86% helical in the pore, 85% at 0.2 μs, and 74% at 1 μs in lipids. As the peptide unfolds and reduces its tilt during the simulation, the peptide-lipid interaction and peptide-water interaction improve at the expense of intramolecular interaction. In the interactions between headgroups and C-termini, ∼20–30% is van der Waals and the rest electrostatic (Table S5 shows the headgroup interactions of each C-terminal residue). Overall, MelP5 appears to be less “happy” as an isolated peptide surrounded by lipids than melittin, and that forces it to stay on the edge of an aqueous pore. The reason for this and for the unfolding at the C-terminus appears to be the relatively unfavorable interaction of the alanine and glutamine residues with the lipid headgroups. The transfer of amino acid side-chain analogs from water to the bilayer interface was calculated to be significantly less favorable for alanine and glutamine than for lysine and arginine (57). The constraints imposed by attachment to a helix could make the interactions even less favorable. This apparently causes the MelP5 helix first to insert more deeply and tilt (presumably to bury the alanine residues) and then to unfold and move these residues into water.

Table 2.

Average Interaction Energies during the Last Microsecond of the MelP5 to Melittin Simulation^a and MelP5 Parallel-Hexamer Simulation^b

	With Water	With Lipids	With Tails	N-Terminus with Headgroups	C-Terminus with Headgroups	Intramolecular
Melittin in pore	−282 ± 23	−274 ± 12	−70 ± 10	−121 ± 21	−61 ± 10	390 ± 7
Melittin in lipids	−223 ± 16	−386 ± 19	−107 ± 10	−193 ± 15	−86 ± 18	395 ± 8
ΔE pore → lipids	+59 ± 28	−112 ± 22	−37 ± 14	−72 ± 26	−25 ± 21	+5 ± 11
MelP5 in pore	−225 ± 29	−292 ± 60	−77 ± 9	−156 ± 58	−29 ± 7	382 ± 2
MelP5 in lipids 0.2 μs	−177 ± 11	−339 ± 25	−117 ± 8	−151 ± 26	−55 ± 6	400 ± 3
ΔE pore → lipids 0.2 μs	+48 ± 31	−47 ± 65	−40 ± 12	+5 ± 64	−26 ± 9	+18 ± 4
MelP5 in lipids 1 μs	−205 ± 28	−411 ± 27	−93 ± 9	−229 ± 31	−58 ± 15	434 ± 10
ΔE pore → lipids 1 μs	+20 ± 40	−119 ± 66	−16 ± 13	−73 ± 66	−28 ± 17	+52 ± 10

In the case of the MelP5 single-lipid simulations, the interaction energy values are averaged over 150–200 ns or 950–1000 ns of the 1-μs trajectory. The “With Water” and “With Lipids” columns show the interaction of each monomer with the solvent and the bilayer as a whole, respectively. The next two columns show the interaction energies between the N-terminus (residues 1–7) and C-terminus (residues 21–26) with the lower and upper headgroup regions, respectively. The last column shows the intramolecular energy of each peptide. The error bars are SDs of block averages. The p-values associated with the pore → lipid differences are all below 0.05. “Melittin in pore” and “in lipids”, and “MelP5 in pore” rows correspond to the average interaction energies during the last microsecond of the MelP5 to melittin simulation and MelP5 parallel hexamer simulation, respectively. The underlining highlights the difference in interaction energies.

With the exception of one of the peptides moving to a surface orientation in the melittin octamer simulation, the dissociated monomers adopt a fully I-state configuration. The transmembrane orientation is also more preferred in the pore by the melittin monomers compared to the MelP5 peptides, which favor more tilted conformations (last microsecond average values of 24 ± 15° and 39 ± 11°, respectively) (Fig. 3). The larger tilt angles for MelP5 could be due to two effects: first, the longer hydrophobic length of this peptide compared to melittin, and second, the relatively unfavorable interactions of the C-terminal region with headgroups compared to melittin (Table 2). A similar observation has been reported in recent coarse-grained melittin and MelP5 simulations (36). The average tilt-angle values for the melittin peptides are in good agreement with those obtained by solid-state NMR in DMPC at a P/L ratio ∼1:10 (58), especially for the peptides that dissociate from the pore in an I-state orientation. The melittin and MelP5 monomers in contact with the pore are able to sample a wider array of conformations by orienting their charged groups toward the water channel and display tilt-angle values similar to those observed in other experimental and computational works (30). The simulations also show that the MelP5 and melittin peptides in contact with the water channel share a similar dynamic behavior.

Melittin peptides are able to diffuse freely through the bilayer either individually or in the dry trimer. The dissociation of peptides—either individually or in small oligomeric assemblies—observed in these simulations support that single melittin peptides could actually adopt a stable pore-independent inserted orientation. The transient restoration of the tetrameric pore in the melittin parallel-hexamer simulation leads us to think that these peptides could comprise a pool of available individual monomers that can either merge into a growing pore or to further stabilize an existing one. The ability to incorporate and lose monomers may account for the dynamic behavior of melittin pores observed in different experimental studies. We should raise the possibility, however, that this behavior may depend on lipid properties and, in particular, membrane thickness. DMPC and 1,2-dimyristoyl-SN-glycero-3-phosphoglycerol membranes are relatively thin and may favor a transmembrane orientation of melittin (59). In thicker membranes, such as 1,2-dioleoyl-SN-glycero-3-phosphocholine, the stability of dry monomers or oligomers in the I-state might be reduced. This issue should be investigated by simulations in thicker membranes.

Conclusions

This work examines pore dynamics and stabilization from initially inserted oligomeric structures for melittin and its more active analog MelP5 in DMPC bilayers. To our knowledge, this is the first study that provides atomistic structural information of MelP5 pores on the microsecond timescale. We observe a reproducible tendency of high oligomeric state pores to break up into smaller ones for melittin but not for MelP5. Interaction-energy analysis shows that this is not due to higher peptide-peptide electrostatic repulsions in melittin. We have computed surprisingly low electrostatic repulsions between individual like-charged amino acid residues. We find that the highly charged C-terminal regions of melittin, rather than repulsing each other, contribute to the formation of intermolecular interactions with neighboring monomers and lipid headgroups that have a direct impact on both pore structure and size. These findings suggest that amino acid side chains are much more complex than spherical ions, and their interactions might be very challenging to reproduce with coarse-grained representations.

The greater stability of MelP5 pores appears to be due to a larger penalty for moving this peptide from the pore to the membrane compared to melittin. In other words, an isolated transmembrane MelP5 helical peptide in DMPC appears less stable than the corresponding melittin peptide. This is also counterintuitive, as MelP5 has a larger hydrophobic length, and appears to be due to a relatively unfavorable interaction of alanine and glutamine residues with the lipid headgroups. More work needs to be done to characterize the affinity for the membrane interface of each amino acid side chain when it is attached to a helix. From published results on side-chain analogs (57) the least favorable side chains at the interface are the acidic ones, glutamate and aspartate. One might then hypothesize that these residues near the C-terminus of melittin might favor pore formation even more strongly than alanine and glutamine. It is noted that a different combinatorial library aimed at discovering pH-dependent pore formers placed a glutamate or aspartate at position 4 of melittin, which should also be in the headgroup region (60). In any case, the equilibrium between pore edge and membrane interior should be considered in the rational design of pore-forming peptides.

Individual melittin monomers, as well as small oligomers, have the ability to diffuse freely through the lipid bilayer in a stable transmembrane orientation. In their ability to do so, these peptides have the potential ability to merge into the preexisting pore to restore a higher oligomeric state. This phenomenon is observed only once in this work (the melittin parallel-hexamer system), but it could be a recurrent event in longer-timescale MD simulations. Ultimately, the transition between pore and individual transmembrane states of these monomers could account for the dynamic behavior of melittin pores observed in distinct experimental setups.

Both melittin and MelP5 stabilize toroidal pores that increase their size with the number of peptides supporting them. Parallel arrangements of the peptides lead to large pores and thus appear more consistent with experimental findings. However, the energetic favorability of this arrangement, which is also counterintuitive in terms of electrostatic interactions, still needs to be established.

Author Contributions

A.P.-A. and T.L designed the research. A.P.-A. performed the simulations. A.P.-A. and T.L. analyzed the data and wrote the article.

Editor: Emad Tajkhorshid.

Supporting Citations

References (61, 62, 63, 64, 65, 66, 67, 68) appear in the Supporting Material.