Could Cardiolipin Protect Membranes against the Action of Certain Antimicrobial Peptides? Aurein 1.2, a Case Study.

The activity of a host of antimicrobial peptides has been examined against a range of lipid bilayers mimicking bacterial and eukaryotic membranes. Despite this, the molecular mechanisms and the nature of the physicochemical properties underlying the peptide-lipid interactions that lead to membrane disruption are yet to be fully elucidated. In this study, the interaction of the short antimicrobial peptide aurein 1.2 was examined in the presence of an anionic cardiolipin-containing lipid bilayer using molecular dynamics simulations. Aurein 1.2 is known to interact strongly with anionic lipid membranes. In the simulations, the binding of aurein 1.2 was associated with buckling of the lipid bilayer, the degree of which varied with the peptide concentration. The simulations suggest that the intrinsic properties of cardiolipin, especially the fact that it promotes negative membrane curvature, may help protect membranes against the action of peptides such as aurein 1.2 by counteracting the tendency of the peptide to induce positive curvature in target membranes.

Introduction

Whenever pathogens are exposed to antibiotic agents, resistance can develop. Indeed, resistance to antibiotic agents that are used clinically has become a global public health problem.[1] Membrane-disrupting antimicrobial peptides have, however, remained effective against bacteria on an evolutionary timescale and thus have the potential to form the basis of a new class of therapeutics.[2] These relatively small peptides are found throughout the animal and plant kingdoms and are vital components of the defense systems of complex multicellular organisms.[2] Membrane-disrupting antimicrobial peptides bind to the surface of microbial membranes and beyond a threshold concentration, induce the formation of transmembrane pores, or otherwise destabilize the membrane. The skin secretions of many amphibians are rich in such peptides. In particular, Australian tree frogs of the genus Litoria secrete a series of pore-forming or membrane-disrupting antimicrobial peptides that are in general active against both Gram-positive and Gram-negative bacteria.[3,4] Amongst them is the peptide aurein 1.2. Aurein 1.2 has 13 amino acids and a net charge of +1e at neutral pH. It is unstructured in aqueous solution but folds into an α-helix upon binding to model membranes.[5−8] Aurein 1.2 is the smallest amphibian peptide to show antibiotic and anticancer activity.[3] Because aurein 1.2 is too short to span a lipid bilayer but is membrane-lytic, it has been generally considered to operate via the so-called “carpet” mechanism whereby the peptide destabilizes membranes in a detergent-like manner leading to micellization. This has been supported by a range of studies, including vesicle leakage experiments followed by fluorescence spectroscopy,[9,10] solid-state nuclear magnetic resonance (NMR) spectroscopy,[7,10] neutron reflectometry,[10] quartz crystal microbalance,[10−13] and dual polarization interferometry[14,15] experiments. Collectively, these experiments show strong interaction between aurein 1.2 and lipid headgroups prior to a rapid aggregation-driven micellization of lipid bilayers beyond a threshold peptide concentration that is itself a function of the lipid composition. Coarse-grained simulations of aurein 1.2 with zwitterionic and anionic lipid bilayers have also suggested that aurein 1.2 might form pore-like structures.[16,17]

The membrane-binding properties of aurein 1.2 have been extensively examined experimentally using a wide range of zwitterionic (phosphatidylcholine and phosphatidylethanolamine) and anionic (phosphatidylglycerol) lipid bilayers.[6,11,13,18−20] Overall, aurein 1.2 appears to interact preferentially with anionic lipids, probably because of its cationic nature.[6,21] For example, the effect of aurein 1.2 on the thermotropic phase behavior of a series of phospholipid bilayers, including 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC), 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine, and 1,2-dimyristoyl-sn-glycero-3-phosphoglycerol (DMPG) bilayers, has been investigated using differential scanning calorimetry. The greatest effect was with a DMPG bilayer, suggesting strong interactions between aurein 1.2 and DMPG.[6] Similarly, mixed DMPC/DMPG bilayers have been found to be more sensitive to lysis by aurein 1.2 than pure DMPC bilayers.[10,11,15] This preferential interplay with anionic lipid bilayers is consistent with the antibiotic and anticancer activity of aurein 1.2, as anionic lipids are commonly found in the outer membrane leaflet of both bacterial and cancer cells.[22−24] However, despite strong interactions of aurein 1.2 with membranes containing anionic lipids, aurein 1.2 is not very potent against Escherichia coli and Staphylococcus aureus or lipid bilayers consisting of either an E. coli lipid extract or lipids mimicking the composition of a S. aureus membrane.[8,14,25] Both these bacteria contain significant amounts of the anionic lipid cardiolipin. Cardiolipin is a vital component of membranes in which electron transport and phosphorylation are coupled, namely, bacterial plasma membranes, chromatophores, chloroplasts, and mitochondria. It has a unique “double-lipid” structure in which two 1,2-diacylglycerol 3-phosphate groups are connected by a central glycerol moiety. The quadruple-chain structure of cardiolipin leads to a high degree of cohesion in the interfacial region of a cardiolipin-containing bilayer and results in an increase in the structural integrity of the bilayer.[26] Although this might explain why membranes containing cardiolipins appear to be less sensitive to the effects of some antimicrobial peptides and antibiotics than are membranes containing other anionic lipids,[26−29] the precise molecular basis for this finding is still unclear.

In this study, the interaction of aurein 1.2 with model membranes containing cardiolipin was investigated using atomistic molecular dynamics simulations. The effect of varying amounts of peptide over a range of temperatures was simulated in order to examine the potential role played by cardiolipin in maintaining the integrity of the membranes.

Theoretical Methods

Simulation Systems

Aurein 1.2 (GLFDI IKKIA ESF-NH2) is a cationic, 13-residue, C-amidated peptide with a net charge of +1e under physiological conditions. Like in a number of other tree frog antimicrobial peptides, the C-terminus of aurein 1.2 is amidated, which is essential for its antimicrobial and anticancer activity.[30] The initial structure of aurein 1.2 was built as an ideal α-helix using PyMOL.[31] The behavior of aurein 1.2 was investigated in a range of systems, as detailed in Table . In the first system (system A), the structure of a single aurein 1.2 molecule was examined in an aqueous environment. In system B, one copy of aurein 1.2 was simulated in the presence of a mixed cardiolipin/phosphatidylglycerol bilayer. The peptide was initially placed in the water phase. The 102-lipid bilayer consisted of a mixture of phosphatidylglycerol lipids (namely, POPG, 2-oleoyl-1-palmitoyl-sn-glycero-3-phosphoglycerol) with 25 mol % of cardiolipin (TOCL, 1,1′,2,2′-tetraoleoylcardiolipin), corresponding to 76 POPG molecules with 26 TOCL molecules. The phosphatidylglycerol component of the bilayer was a racemic mixture of 2-oleoyl-1-palmitoyl-sn-glycero-3-phospho-l-(1-glycerol) (l-POPG) and 2-oleoyl-1-palmitoyl-sn-glycero-3-phospho-d-(1-glycerol) (d-POPG). The lipids cardiolipin and phosphatidylglycerol are present in a broad range of Gram-positive and Gram-negative bacterial membranes but tend to be found in higher concentrations in Gram-positive bacteria. The 3:1 POPG/TOCL bilayer composition used in this study was designed to mimic a Gram-positive bacterial membrane. The relative phosphatidylglycerol-to-cardiolipin ratio can vary greatly amongst bacteria (about 1:1 in S. aureus and Streptococcus pneumoniae, 2.5:1 in Bacillus cereus, and 18:1 in Bacillus subtilis).[24,32] The cardiolipin content in membranes also changes in space (cardiolipin-enriched domains) and time (stage in the life cycle).[24,33,34] Furthermore, bacteria can contain significant amounts of O-aminoacylated glycerophospholipids, especially O-(l-lysyl)phosphatidylglycerol lipids in S. aureus.[35,36] Although O-(l-lysyl)phosphatidylglycerol lipids have been shown to promote resistance to antibiotics and cationic antimicrobial peptides in S. aureus(37,38) and other bacteria[39,40] by changing the overall charge of the membrane surface, its protective effect has been primarily ascribed to a stabilization of the membrane, thereby hampering disruption or perturbation of the membrane induced by antibiotics and cationic antimicrobial peptides. O-(l-Lysyl)phosphatidylglycerol lipids would then have only a minimal effect on the binding of antibiotics and cationic antimicrobial peptides to the membrane.[35,41,42] Therefore, although the lipid composition used in the simulations did not include O-(l-lysyl)phosphatidylglycerol, this is not expected to affect the binding of aurein 1.2 onto the mixed POPG/TOCL bilayer. The initial configuration of the bilayer was constructed using PACKMOL[43] and equilibrated for 56 ns before the addition of a peptide. Cl– and Na+ counterions were included to neutralize each charge in the systems. All systems comprised sufficient water molecules to ensure a fully hydrated state.

Table 1

Overview of the Systems Simulateda

system label	number of simulations	number of peptides	number of lipids	P/L	time (ns)b	T (K)	binding restrictionc	initial peptide structure
A	3	1			100	310		α-helical
B	3	1	26 TOCL	1:100	100	298	no	α-helical
			76 POPG
C1	1	10	104 TOCL	1:41	80	298	yes	nonhelical
			304 POPG
C2	1	10	104 TOCL	1:41	80	318	yes	nonhelical
			304 POPG
C3	1	10	104 TOCL	1:41	80	338	yes	nonhelical
			304 POPG
C4	1	20	104 TOCL	1:20	40	298	no	nonhelical
			304 POPG
C5	1	20	104 TOCL	1:20	40	338	no	nonhelical
			304 POPG
D1d	1	10	104 TOCL	1:41	20 (20)	298	yes	nonhelical
			304 POPG
D2	1	15	104 TOCL	1:27	75 (55)	298	yes	nonhelical
			304 POPG
D3	1	30	104 TOCL	1:14	100 (25)	298	yes	nonhelical
			304 POPG
D4	1	40	104 TOCL	1:10	200 (100)	298	yes	nonhelical
			304 POPG
E1	1	0	104 TOCL	0	40	298
			304 POPG
E2	1	0	104 TOCL	0	40	298
			304 POPG

T, temperature; P/L, peptide-to-lipid ratio; POPG, 2-oleoyl-1-palmitoyl-sn-glycero-3-phospho-rac-(1-glycerol); and TOCL, 1,1′,2,2′-tetraoleoylcardiolipin.

The time in brackets for simulations D1–D4 indicates the simulation time for the particular systems excluding the cumulated time over previous simulations.

The peptides were either left free to bind to either leaflet of the lipid bilayer or allowed to bind to only one of the two leaflets by position-restraining a layer of water molecules between the peptides and the periodic image of the lipid bilayer along z, thereby stopping the peptides from diffusing freely to the other leaflet.

D1 corresponds to the first 20 ns of C1.

Systems A and B were simulated three times each starting with different initial velocities. Each simulation was 100 ns in length. In the case of system B, the initial configuration of aurein 1.2 was taken from one of the simulations of system A. Systems C and D contained larger numbers of aurein 1.2. In order to accommodate the larger number of peptides, the TOCL/POPG bilayer was duplicated along the x and y axes (the two directions defining the plane of the bilayer), resulting in a bilayer of 408 lipids.

In systems C1, C2, and C3, 10 peptides are bound to the same leaflet of the bilayer. Simulations were performed at three different temperatures: 298, 318, and 338 K corresponding to systems C1, C2, and C3, respectively. At the start of the simulations, the peptides were placed in the water phase. To ensure all peptides were bound to the same leaflet, the positions of a layer of water molecules were restrained by applying a harmonic potential to each water oxygen with a force constant of 100 kJ mol–1 nm–2 to make a barrier, preventing the diffusion of peptides toward the other leaflet. The peptides were placed between this barrier and the bilayer surface. The position restraints on the water molecules were removed once all peptides were bound to the bilayer surface. In systems C4 and C5, 20 copies of aurein 1.2 were placed in the water phase. In these cases, the peptide was free to bind to either of the two leaflets of the lipid bilayer. The simulations were performed at 298 K (system C4) and 338 K (system C5).

In D, the number of peptides was progressively increased from 10 (system D1) to 40 (system D4). System D1 consisted of the first 20 ns of simulation of system C1. After 20 ns, five peptides were added to system D1, leading to system D2 (15 peptides). After another 55 ns, a further 15 peptides were added to yield system D3 (30 peptides). Then, after 25 ns, 10 more peptides were added to give system D4 (40 peptides). Systems D3 and D4 were simulated for 25 and 100 ns, respectively. The cumulated simulation time over systems D1, D2, D3, and D4 was 200 ns. The peptides in D1 to D4 were restricted to binding only to one leaflet by using a layer of position-restrained water molecules as described previously. Note, in systems C and D, the level of hydration corresponded to 70–100 water molecules per lipid. The configuration of each extra copy of aurein 1.2 was selected randomly from the simulation of aurein 1.2 in water (system A). The secondary structure was a mixture of partially α-helical and nonhelical structures.

Finally, two systems (E1 and E2) were derived from snapshots of the simulation of systems D1 (E1) and D2 (E2) at 20 and 60 ns, respectively. These times correspond to the cumulated time over systems D1 and D2. All copies of the peptide were removed. Position restraints were applied to all nonwater atoms, and the system progressively relaxed by gradually decreasing the force constant from 500 kJ mol–1 nm–2 to 0 over 5 ns. Systems E1 and E2 were then simulated for 40 ns.

Experimentally, vesicle leakage has been observed for peptide-to-lipid ratios (P/L) in the range of 1:100 to 1:5 for phosphatidylcholine and mixed phosphatidylcholine/phopshatidylglycerol vesicles.[9,10] Given that P/L ranged from 1:100 to 1:10 in systems B, C, and D, the peptide concentrations used was favorable for membrane disruption to occur in the simulations. Note, aurein 1.2 has been shown to inhibit the growth of a range of Gram-positive bacterial strains at concentrations within 1–100 μg mL–1.[8,44,45] Variations in these concentrations between bacterial genera and strains may be due to specific membrane or cell wall components promoting or hindering the lytic activity of aurein 1.2 that are yet to be identified.

Simulation Parameters

All simulations were performed using the GROMACS simulation package version 3.3.3.[46] The united-atom GROMOS 54a7 force field[47] was used to describe aurein 1.2. Force-field parameters for POPG and TOCL were derived from the GROMOS 54a7 force field for lipids modified by Kukol.[48−50] The simple-point-charge (SPC) water model was used to describe the solvent water.[51] Periodic boundary conditions and a minimum-image convention were used in all three directions. All simulations were performed at constant temperature and pressure. The temperature and pressure were maintained using the weak-coupling method of Berendsen[52] with time constants of 0.1 and 4 ps for the temperature and pressure, respectively. The simulations of the isolated peptides in water (system A) were performed at a constant temperature of 310 K. The simulations of the other systems that include a lipid bilayer were performed either at 298, 318, or 338 K as indicated in Table . To our knowledge, the phase transition temperature Tm from a gel to a liquid-crystalline phase has not been determined for a mixed POPG/TOCL bilayer. However, experimental studies of binary mixtures of the related cardiolipin molecule TMCL (1,1′,2,2′-tetramyristoylcardiolipin) with 1,2-dimyristoyl-sn-glycero-3-phospho-rac-(1-glycerol) (DMPG) and 1,2-dipalmitoyl-sn-glycero-3-phospho-rac-(1-glycerol) (DPPG) indicated that the lipid species mixed well, almost ideally, and that the observed values of Tm in mixtures containing 20–30 mol % TMCL were within a temperature range defined by the Tm values of the lipid species and closer to that of a pure DMPG or DPPG bilayer.[53,54] Therefore, given that for both POPG and TOCL, Tm < 273 K,[55−59] it is reasonable to assume that a POPG/TOCL bilayer with 25 mol % TOCL is in a fluid phase at the temperature examined. In addition, an equimolar mixture of POPG and TOCL was shown to form a stable, homogeneous lamellar phase.[60] The temperatures of the solute and solvent were independently coupled to the temperature bath. A reference pressure of 1 bar was used. For the simulation of the systems comprising a single peptide (systems A and B), a 2 fs time step was used. In all other simulations (systems C1, C2, C3, C4, C5, D1, D2, D3, D4, E1, and E2), a time step of 4 fs was used. This was achieved by a combination of replacing some hydrogens by virtual interaction sites (dummy atoms) and increasing the mass of others.[61] This removes high-frequency motions allowing a larger time step to be used without affecting the thermodynamic properties of the systems significantly. The lengths of all covalent bonds within the peptide and lipid molecules were constrained using the LINCS algorithm.[62] The geometry of the SPC water molecules was constrained using SETTLE.[63] Lennard-Jones and electrostatic nonbonded interactions were described using a twin-range method. Interactions within the shorter range cutoff of 0.8 nm were calculated every step, whereas interactions within the longer range cutoff of 1.4 nm were updated every second time step together with the generation of the neighbor list. To correct for the truncation of interactions beyond the 1.4 nm long-range cutoff, a reaction-field correction was applied using a relative dielectric constant of 78.[64]

Analysis

The evolution of the secondary structure of aurein 1.2 in the simulations was determined using the program STRIDE.[65] The helicity of aurein 1.2 was calculated as the percentage of amino acids assigned to be in an α-helix by STRIDE with respect to the total number of amino acids in the peptide (13). The average probability for an amino acid to fold into an α-helix during the simulations is referred to as the helix propensity.

The propensity of each amino acid to interact with a lipid bilayer (contact probability), was estimated by calculating the frequency that any atom of a given amino acid was within 0.3 nm of any lipid atom.

The relative solvent accessible surface area (SASA) of each residue in aurein 1.2 was calculated using the method of Lee and Richards with a 0.14 nm solvent probe.[66,67]

To evaluate the extent to which the lipid bilayer underwent deformation in the simulation, the ratio between the area a of the simulation box within the plane of the bilayer (xy plane) at time t during the simulation and the corresponding area a0 in the initial frame of the simulation at t = 0 was calculated. For the simulations of systems C2, C3, C4, C5, D1, D2, D3, D4, E1, and E2, the value of a0 was taken from the initial frame of the simulation of system C1. If a/a0 ≈ 1, then the lateral area of the bilayer is unchanged and the surface of the bilayer stayed flat. In contrast, a/a0 < 1 indicates that either the bilayer remained planar but the lipids were packed more densely or the bilayer experienced bending, buckling, rippling, or infolding.

The interaction of aurein 1.2 with a lipid species was assessed by calculating the distance r of carbon α of Phe3 in aurein 1.2 with every lipid of a given species (TOCL, d-POPG or l-POPG) within a radius rmax = 1.5 nm, taking periodic images into account. The position of the lipids was taken as the position of the phosphorus atom in d-POPG and l-POPG, and that of the phosphorus atom in TOCL that lay closer to the Cα of Phe3. Then, the frequency f(r) for aurein 1.2 to be at a distance r ≤ rmax from each lipid species throughout the simulation was binned usingwhere n(r) is the number of occurrences in the simulation when r was within a bin r1 < r ≤ r2. The bin size r2 – r1 was 0.05 nm.

Results

Structure of Aurein 1.2 and Interaction with a Lipid Bilayer

Circular dichroism, NMR, and infrared spectroscopy experiments have shown that aurein 1.2 is primarily unstructured in aqueous solution but readily folds into an amphipathic α-helix in a solution containing d3-trifluoroethanol (70% v/v in water) or in a membrane-mimetic environment.[3,5−8,68] To examine the effect of the environment on its structure, aurein 1.2 was simulated in bulk water and in the presence of a TOCL/POPG model membrane (systems A and B, respectively; Table ). Note, in the absence of a peptide, the TOCL/POPG lipid bilayer relaxed to a state in which the area per lipid AL calculated for each lipid species using a grid-based approach[69] was 0.85, 0.86, and 1.02 nm2 for d-POPG, l-POPG, and TOCL, respectively. Despite the absence of experimental data on a fluid-phase POPG/TOCL bilayer, this is consistent with AL values determined experimentally for fluid-phase TOCL and TMCL bilayers (1.298 and 1.04 nm2)[70,71] and mixtures in which TMCL was found to increase the area per lipid of DMPC in a DMPC/TMCL bilayer. Specifically, it was estimated that AL was 0.74 nm2 in a 4:1 DMPC/TMCL bilayer, whereas AL = 0.64 nm2 in a fluid-phase DMPC bilayer.[70] Given that AL = 0.647 nm2 for a fluid-phase POPG bilayer,[72]AL is likely to be around 0.75–0.90 nm2 in a 3:1 POPG/TOCL bilayer.

In bulk water and in the presence of a TOCL/POPG bilayer, the peptide was initially α-helical. In aqueous solution, aurein 1.2 transitioned between a helical and a nonhelical structure multiple times during the 100 ns simulation. The average helicity was 47% over the three simulations. In the presence of a TOCL/POPG bilayer, the peptide is bound to the lipid bilayer within the first 4 ns in all simulations and remained associated with the membrane, thereafter. Like in the simulations in solution, the helical content of aurein 1.2 tended to vary between 0 and 30–50%. This is illustrated in Figure A which shows the time evolution of the helicity of aurein 1.2 during one of the three simulations of system B. The region of the peptide most likely to adopt an α-helical structure was composed of residues Phe3 to Lys8, which showed an average helix propensity of greater than 20% (inset in Figure A). The average probability of contact between each amino acid in aurein 1.2 and the lipid bilayer over all three simulations is displayed in Figure A. Note, a contact between an amino acid and the TOCL/POPG lipid bilayer was defined when any atom of this amino acid was within 0.3 nm of any atom in a lipid molecule. From Figure A, it is clear that the residues that were associated with the lipid bilayer more than 50% of the time (Gly1, Leu2, Phe3, and Phe13) are hydrophobic. This is in line with previous experimental and coarse-grained simulation studies that identified Phe3 and Phe13 as the key residues that anchor aurein 1.2 within a range of lipid bilayers.[13,73] A helical wheel projection of the structure of aurein 1.2 is shown in Figure B. As can be seen, Leu2, Phe3, and Phe13 lie on the same face of aurein 1.2 when it adopts an α-helical structure (Figure B). However, despite its clear amphipathic character, in the simulations, aurein 1.2 was on average only about 50% α-helical, and a range of alternative binding modes of aurein 1.2 on the TOCL/POPG bilayer was also observed. Indeed, residues Ile5, Ile6, Ile9, and Ala10 that form the central part of the peptide and in the simulations had the highest helical propensity, and did not interact strongly with the membrane. In fact, the contact probability of Ile5, Ile6, and Ala10 was less than that of Lys7 and Lys8 despite lying on the opposite side of the helical wheel projection. The relative solvent-accessible surface area (SASA) was calculated for each amino acid and with respect to the simulation of aurein 1.2 in solution in system A (Figure C). It is interesting to note that for all residues except Phe3, there was relatively little change in the extent to which residues were solvent-exposed whether the peptide was free in solution or bound to the lipid bilayer. However, Phe3 was clearly involved in interactions with lipids and showed a marked decrease in solvent accessibility in all simulations of system B (average relative SASA of 36%).

Figure 1

Time evolution of the helicity of aurein 1.2 in the presence of a TOCL/POPG bilayer in the simulations of (A) a single peptide (system B), (B) 10 copies of the peptide (systems C1, C2, and C3), and (C) 10–40 copies of the peptide (systems D1, D2, D3, and D4). In panel A, the inset shows the probability of each residue along the aurein 1.2 sequence to lie in an α-helix. In panel B, the helicity is averaged over all 10 peptides. In panel C, the helicity corresponds to the average over the 10 peptides that were initially present in system D1. The green vertical dashed lines correspond to the times when new copies of aurein 1.2 were added to the system simulated.

Figure 2

Structural properties of aurein 1.2 and its interaction with a TOCL/POPG bilayer in the simulation of system B. (A) Probability of contact between each residue along the aurein 1.2 sequence and any lipid molecule in a TOCL/POPG bilayer. (B) A helical wheel representation of aurein 1.2. Hydrophobic, polar, positively charged, and negatively charged residues are shown in gray, pink, red, and blue, respectively. (C) Relative SASA of each residue along the aurein 1.2 sequence.

Deformation of a Lipid Bilayer due to Aurein 1.2 Adsorption

To examine the effect of binding of multiple copies of aurein 1.2 on a lipid bilayer, 10 copies of aurein 1.2 were simulated in the presence of a TOCL/POPG bilayer at 298, 318, and 338 K (systems C1, C2, and C3, respectively; Table ). Note, all peptides were restrained to bind to the same leaflet of the bilayer. In all cases, the peptides were bound to the bilayer within 20 ns and remained bound throughout the rest of the simulations. The binding of the peptide to the lipid bilayer at 298 K was associated with an increase in the average helical content in the peptides. This is indicated by the black line in Figure B, which corresponds to the time evolution of the average helicity of aurein 1.2 in the simulation of system C1. The red and blue lines in Figure B correspond to the simulations at 318 and 338 K and show a loss of helicity with increasing temperature. The average helicity calculated over the 10 peptides from 20 ns on for systems C1, C2, and C3 was 27, 21, and 3%, respectively. Snapshots of the structure of the peptide-bound TOCL/POPG bilayer at the end of the simulations of systems C1, C2, and C3 are shown in Figure . As can be seen, the binding of 10 copies of aurein 1.2 induced significant buckling of the lipid bilayer, especially at 338 K (system C3). As an estimate of the deviation of the bilayer surface from planarity, the change in the lateral area of the simulation box a/a0 was calculated. Figure A displays the variations of a/a0 throughout the simulations of systems C1 and C3. In both simulations, the area of the simulation box decreased significantly. This is especially evident in the case of system C3 (a/a0 ≈ 0.85 after 80 ns). In system C1, a/a0 stabilized at about 0.93. These values are consistent with the extent of bending and curvature observed in Figure . Note, a layer of water molecules was position-restrained initially to ensure the initial bending of the peptides to the same leaflet. The position restraints were removed after all the peptides were bound to the leaflet, that is, within 20 ns of simulation for all systems. It is unlikely that the deformation undergone by the lipid bilayer was caused by these temporary position restraints as, for example, the bilayer remained flat (a/a0 ≈ 1) in simulation C3 for the first 30 ns simulation. Simulations of systems C4 and C5 at 298 and 338 K, respectively, were performed as controls. The peptide was free to bind to either of the two leaflets and, in this case, only negligible curvature was observed, even at 338 K (Figure ). In both simulations, a/a0 remained close to 1 (Figure A). Indeed, for system C5, a/a0 increased slightly to about 1.02–1.03 on average as a result of an expansion of the lipid bilayer at high temperature. This clearly indicates that the distortion of the lipid bilayer was associated with the asymmetric binding of aurein 1.2. Despite the high degree of buckling of the bilayer in system C3, the structural integrity of the bilayer was maintained. The peptides neither aggregated nor penetrated into the bilayer in any of the simulations (Figure ). The amino acids that were in close contact with the lipids at 298 K in the simulation of system C1 (contact probability greater than 50% in Figure ) were similar to those observed in the simulation of a single peptide bound to a TOCL/POPG bilayer in system B (Figure A), that is, Gly1, Leu2, Phe3, Asp4, Lys7, Lys8, and Phe13. Again, the contact probability of Phe3 was noticeably higher (87%) than for the other residues for which the contact probability was in the range of 50–64%. At higher temperatures in the simulations of systems C2 (318 K) and C3 (338 K), the contact probability increased for all residues, possibly because the peptides were less structured, thereby exposing all the residues to the bilayer. Figure displays histograms of the frequency f(r) for d-POPG, l-POPG, and TOCL in the simulations of systems C1, C2, and C3 (panels A, B, and C, respectively) to lie at a distance r ≤ 1.5 nm from a peptide taking the position of the carbon α of Phe3 in the peptide as a reference. Phe3 was chosen as a reference because it was identified as a key membrane-anchoring residue in all simulations. Note, f(r) was averaged over all 10 copies of aurein 1.2. As can be seen, there were more POPG molecules than TOCL molecules in the immediate vicinity (r ≤ 0.7 nm) of aurein 1.2 in all simulations. This suggests that aurein 1.2 may interact preferentially with POPG over TOCL. There was no clear distinction between the two enantiomers of POPG.

Figure 3

Side (top row) and top (bottom row) views of the structure of a TOCL/POPG bilayer in the presence of 10 copies of aurein 1.2 at the end of the simulations of systems C1 (panels A and D), C2 (panels B and E), and C3 (panels C and F) (t = 80 ns). The dashed lines indicate the edge of the unit cell. Aurein 1.2 is shown in green, POPG in cyan, and TOCL in pink. Oxygen atoms in POPG and TOCL are drawn as blue and purple spheres, respectively.

Figure 4

Relative change in the lateral area of the simulation box (a/a0) in (A) the simulations of systems C1, C3, C4, and C5, (B) the simulations of systems D1, D2, D3, and D4, and (C) the simulations of systems E1 and E2.

Figure 5

Lateral view of a TOCL/POPG bilayer in the presence of 20 copies of aurein 1.2 at the end of the simulation of system C5 (t = 40 ns). The dashed lines indicate the edge of the unit cell. Aurein 1.2 is shown in green, POPG in cyan, and TOCL in pink. Oxygen atoms in POPG and TOCL are drawn as blue and purple spheres, respectively.

Figure 6

Average probability of contact between each residue along the aurein 1.2 sequence and any lipid molecule in a TOCL/POPG bilayer in the simulation of systems B1, B2, and B3. The contact probability for each residue is average over the 10 copies of aurein 1.2.

Figure 7

Histograms of the frequency f(r) of the distances between the α carbon of Phe3 in aurein 1.2 and the phosphorus atom in d-POPG and l-POPG molecules or the closer of the two phosphorus atoms in TOCL in the simulations of systems (A) C1, (B) C2, and (C) C3.

Effect of Increasing Asymmetric Binding of Aurein 1.2

The effect of the binding of aurein 1.2 to a TOCL/POPG bilayer was further investigated by gradually increasing the concentration of aurein 1.2, starting from 10 copies of the peptide in system D1 (P/L = 1:41) to 15 in system D2 (P/L = 1:27), 30 in system D3 (P/L = 1:14), and finally 40 copies in system D4 (P/L = 1:10) (Table ). As noted above, the initial binding of 10 aurein 1.2 peptides induced only limited curvature at 298 K. However, as the number of peptides bound to the bilayer was increased (systems D2, D3, and D4), the degree of buckling of the bilayer increased to the point that the bilayer formed hairpin-like structures. Snapshots of the systems at the end of the simulations are depicted in Figure . In all simulations, the peptides remained at the surface of the bilayer. The increasing deformation of the TOCL/POPG bilayer is also shown by the variation of a/a0 in Figure B. Each increase in the concentration of aurein 1.2 on the surface of the bilayer led to a further reduction in the lateral area of the bilayer. Interestingly, Figure shows that aurein 1.2 was bound to the bilayer uniformly, that is, it did not bind preferentially to regions of positive or negative curvature but, instead, to regions where the extent of curvature was greater. Removing the peptides (systems E1 and E2) resulted in a commensurate reduction of the deformation of the bilayer with the bilayer either becoming completely flat in the case of system E1 or retaining some curvature after 40 ns in the case of system E2 (Figure ). This is consistent with the variations of a/a0 illustrated in Figure C. a/a0 converged to close to 1 within 10 ns in the case of system E1, whereas the relaxation of system E2 was slower with a/a0 reaching 0.9 by the end of the simulation. The change in peptide helicity as a result of adding more copies of aurein 1.2 was also examined. The average helicity of the 10 peptides that were included initially in system D1 was calculated during the combined 200 ns of simulation of systems D1, D2, D3, and D4 (Figure C). The vertical dashed lines in Figure C mark the times when more copies of aurein 1.2 were added, corresponding to systems D2, D3, and D4. Note, the variations in helicity in the first 20 ns of the simulation of systems C1 (Figure B) and D1 are identical as they correspond to the same simulation. Despite a spike following the first increase in peptide concentration, the average helicity remained stable fluctuating between 17 and 25%. Neither the number of peptides nor the extent of curvature of the lipid bilayer seemed to affect the helical content of aurein 1.2.

Figure 8

Structure of a TOCL/POPG bilayer to which multiple copies of aurein 1.2 were bound in (A) system D1 (10 peptides; t = 20 ns), (B) system D2 (15 peptides; t = 75 ns), (C) system D3 (30 peptides; t = 100 ns), and (D) system D4 (40 peptides; t = 200 ns). Aurein 1.2 is shown in green, POPG in cyan, and TOCL in pink. Oxygen atoms in POPG and TOCL are drawn as blue and purple spheres, respectively.

Figure 9

Structure of a TOCL/POPG bilayer at the end of the simulation of (A) system E1 and (B) system E2. POPG is shown in cyan and TOCL in pink. Oxygen atoms in POPG and TOCL are drawn as blue and purple spheres, respectively.

Discussion

Because of their high potency and selectivity against given cell types, antimicrobial peptides have attracted wide interest as potential next-generation antibiotics. However, the molecular details underlying cell specificity and the mechanism leading to membrane disruption have remained unclear. The interaction of aurein 1.2 with negatively charged lipid bilayers consisting of phosphatidylglycerol only or mixed with phosphatidylcholine or phosphatidylethanolamine has already been investigated extensively.[6,9−11,13−15,17,19−21,25] Therefore, this study focused on how the presence of cardiolipin in a phosphatidylglycerol bilayer could mitigate or suppress the membrane-disruptive activity of the antimicrobial peptide aurein 1.2. Aurein 1.2 is a short peptide (13 amino acids). Antimicrobial peptides that are too short to span the membrane pores have been assumed to preferentially self-assemble on to the membrane surface and to lyse the membrane in a detergent-like manner.[74] This mechanism, referred to as the carpet model or the detergent-like model can account for the general features observed experimentally, but the mechanism of micellization of membranes at the atomic level has yet to be determined. It has been proposed that antimicrobial peptides may trigger the formation of nonlamellar (hexagonal and cubic) lipid phases, alter lipid packing, change the distribution of lamellar-forming and nonlamellar-forming lipids, or simply neutralize negatively charged lipids.[75] The effect of antimicrobial peptides on the morphology of membranes has alternatively been proposed to be induced by the strain imposed by the insertion of peptides that have a wedge-like structure into the membrane interface, thereby perturbing the lipid packing and order within membranes.[76]

In a previous simulation study, the structure of aurein 1.2 along that of three other cationic antimicrobial peptides from Australian tree frogs, namely, citropin 1.1, maculatin 1.1, and caerin 1.1, was investigated in the presence of a series of membrane environments (micelle, planar lipid bilayer, and lipid bilayer containing a pore).[74] It was found that the peptides lost their α-helical structure when bound to a planar DMPC membrane but remained helical when bound to a region of high local positive curvature, such as a preformed toroidal-shaped pore in a 2-oleoyl-1-palmitoyl-sn-glycero-3-phosphocholine (POPC) bilayer, indicating that the helicity of the peptides depended on the matching of their intrinsic curvature with that of the target lipid bilayer. In that case, the helicity of aurein 1.2 was around 50%,[74] in agreement with estimates from the experiment.[5,6,68] In contrast, in the simulations of systems C1 and D1–D4, the helicity was in general below 25%. The lower helical content suggests that the interaction between aurein 1.2 and the POPG/TOCL bilayer used in this work may prevent aurein 1.2 from adopting a fully α-helical structure but it is important to note that the helical content and the lytic activity of aurein 1.2 on different model membranes are not directly correlated.[8]

In the simulations presented here, aurein 1.2 was allowed to interact with a highly negatively charged lipid bilayer containing cardiolipin. Given that aurein 1.2 has been shown to exhibit a greater affinity toward anionic than zwitterionic lipid bilayers[6,10,25] and the concentration in aurein 1.2 (peptide-to-lipid ratio P/L between 1:100 and 1:10) was within a range consistent with vesicle leakage experimentally,[9,10] the POPG/TOCL mixture was expected to be sensitive to the action of aurein 1.2. In all simulations, aurein 1.2 was bound to the lipid bilayer, with Phe3 and Phe13 playing an important role in anchoring the peptide to the bilayer, in line with previous studies.[13,73] Phe3 seemed to interact preferentially with POPG than with TOCL. This is interesting as although the potential for the binding of the peptide to induce clustering of POPG was not examined in the present study because of statistical limitations, the finding that Phe3 was preferentially surrounded by POPG is consistent with results from small-angle neutron scattering experiments that suggested that aurein 1.2 caused the redistribution and clustering of DMPG lipids in a DMPG/DMPC lipid bilayer.[20]

In none of the simulations was the lipid bilayer disrupted, even at a high peptide-to-lipid ratio of 1:10 (system D4), a concentration at which dye leakage from vesicles composed of an equimolar POPG/POPC mixture has been observed experimentally.[9] Instead, aurein 1.2 induced a high degree of curvature of the lipid bilayer (see Figure D). Indeed micellization of the lipid bilayer may require longer timescales. Furthermore, the size of the system and the periodic boundary conditions may also contribute to an artificial stabilization of the hairpin-like structures induced by aurein 1.2, preventing them from evolving toward micelles. However, it is likely that properties intrinsic to cardiolipin, namely, its structure and its surface charge density, played a key role in reducing the susceptibility of the POPG/TOCL bilayer to aurein 1.2. Cardiolipin consists of two diacylphosphatidate moieties linked together through a glycerol headgroup. The relatively small size of the polar headgroup compared to the volume of the quadruple-chained hydrophobic region promotes the formation of nonlamellar lipid phases, in particular, the inverted hexagonal phase (HII), depending on the pH and the concentration of monovalent or polyvalent cations.[77,78] Cardiolipin increases negative curvature in a lipid bilayer[79] and has been shown to form microdomains in regions of high intrinsic negative curvature in E. coli.[80,81] In short, TOCL and aurein 1.2 are expected to induce opposing effects with respect to the curvature of the lipid bilayer. Thus, in the simulations, TOCL is predicted to work against the membrane-disrupting action of the peptide, even at high concentrations. The impairment of the lytic activity of the antimicrobial peptides magainin 2, polybia-MP1, LL-37, and ΔM2 by negative curvature-inducing lipids has also been reported.[27,28,82] Specifically, phospholipid bilayers that contain HII phase-promoting lipids such as cardiolipin and phosphatidylserine require higher peptide-to-lipid ratios than phosphatidylglycerol for these peptides to form transmembrane pores. Importantly, the buckling resulting from the interaction of aurein 1.2 with the lipid bilayer was reversible. Indeed, after the removal of the peptides, the lateral area of the simulation box relaxed back to its value prior to the binding of the peptides and the surface of the TOCL/POPG lipid bilayer flattened (Figures C and 9). This demonstrates that the deformation of the bilayer in the simulations was caused and maintained by the binding of peptide. Note, although the bilayer was still visibly distorted at the end of the simulation of system E2 (Figure B), the progressive and steady increase of a/a0 simply suggests that longer timescales are required for the system to relax fully.

Furthermore, it has been shown experimentally that the motion and the conformational flexibility of the glycerol headgroup in cardiolipins are very restricted compared to other phospholipids.[26,83] Interlipid interactions involving hydrogen bonding between the headgroups are also limited in cardiolipins, potentially leaving the headgroup more exposed to water, ions, and other solutes such as peptides.[26,71,79] Aurein 1.2 that bears two positively charged lysine residues and a net charge of +1e at physiological pH bound strongly to the surface of a TOCL/POPG bilayer as indicated by the high probability of contacts between the lipids and the majority of the amino acids in the simulations (Figures A and 6). The propensity of aurein 1.2 to fold into an α-helix in these simulations with cardiolipin is lower than in previous computational and experimental work without cardiolipin.[5,6,68,74] Overall, the results of the simulations presented here are consistent with experimental observations that aurein 1.2 is much less potent against E. coli and S. aureus model membranes,[14,25] as compared to bilayers containing phosphatidylglycerol lipids but not cardiolipin.[6,9−11,15,21]

Conclusions

In this study, the interaction of aurein 1.2 with an anionic, cardiolipin-containing lipid bilayer was examined using atomistic molecular dynamics simulations. Although aurein 1.2 has been shown to be potent against anionic lipid bilayers, the effect of the presence of cardiolipin was unclear. In general, aurein 1.2 is bound to the lipid bilayer at various peptide-to-lipid ratios. However, in none of the simulations was aurein 1.2 able to destabilize or disrupt the lipid bilayer, even at high peptide concentrations. Instead, the binding of aurein 1.2 was associated with a high degree of buckling of the lipid bilayer that was reversible in the conditions tested. The simulations suggest that the ability of aurein 1.2 to destabilize membranes by inducing positive curvature in them is opposed by the intrinsic properties of cardiolipin, namely, its structural rigidity in the interfacial region of membranes, the solvent exposure of its charged phosphate groups, and its tendency to promote negative membrane curvature. The presence of cardiolipin in bacterial membranes may therefore be associated with potential resistance mechanisms against membrane curvature-dependent antimicrobial peptides.