The membrane activity of the antimicrobial peptide caerin 1.1 is pH dependent

Abstract

Antimicrobial peptides are an important class of membrane-active peptides that can provide alternatives or complements to classic antibiotics. Among the many classes of AMPs, the histidine-rich family is of particular interest since they may induce pH-sensitive interactions with cell membranes. The AMP caerin 1.1 (Cae-1), from Australian tree frogs, has three histidine residues, and thus we studied the pH dependence of its interactions with model cell membranes. Using NMR spectroscopy and molecular dynamics simulations, we showed that Cae-1 induced greater perturbation of the lipid dynamics and water penetrations within the membrane interior in an acidic environment compared with physiological conditions. Using ³¹P solid-state NMR, the packing, chemical environment, and dynamics of the lipid headgroup were monitored. ²H solid-state NMR showed that Cae-1 ordered the acyl chains of the hydrophobic core of the bilayer. These results supported the molecular dynamics data, which showed that Cae-1 was mainly inserted within the lipid bilayer for both neutral and negatively charged membranes, with the charged residues pulling the water and phosphate groups inward. This could be an early step in the mechanism of membrane disruption by histidine-rich antimicrobial peptides and indicated that Cae-1 acts via a transmembrane mechanism in bilayers of neutral and anionic phospholipid membranes, especially in acidic conditions.

Significance

Histidine-rich antimicrobial peptides (AMPs) are an important class of membrane-active peptides that may induce pH-sensitive interactions with cell membranes. The Australian tree frog AMP caerin 1.1 (Cae-1) has three histidine residues, which possibly modulate the interactions with eukaryotic and prokaryotic membrane systems. NMR and molecular dynamics simulations showed that Cae-1 induced greater perturbation of the lipid dynamics and water penetrations within the membrane interior in an acidic environment compared with physiological conditions. Cae-1 was mainly inserted within the lipid bilayer structure of both neutral and negatively charged membranes, with the charged residues pulling the water inward, which could be an early step in the mechanism of membrane disruption by histidine-rich AMPs.

Introduction

Antimicrobial peptides (AMPs) are a class of membrane-active peptides that interact strongly with the lipid membrane of prokaryotic cells (1,2). The main driving force for the membrane-targeting activity of AMPs is the electrostatic interaction between negatively charged membrane lipids and positively charged amino acids. Indeed, and in particular for bacteria, prokaryotic cell membranes display a net negative charge due to a significant content of anionic phospholipids, such as phosphatidylglycerol (PG) (3,4), compared with eukaryotic cells, which are mainly made of the zwitterionic phospholipid phosphatidylcholine (PC). Cationic AMPs possess a net positive charge due to a high number of the basic residues lysine and arginine but also from the amino acid histidine, which has an imidazole side chain with a pKa of ca. 6 (5). Thus, histidine-containing AMPs can carry an additional positive charge per histidine residue at moderately acidic pH environments. Several histidine-containing AMPs have been shown to change their molecular interactions with lipid membranes depending on pH. Native AMPs, such as the piscidins (6,7), GAD (8), and in-silico-designed AMPs, such as C18G (9) and AHH (10), have been studied at physiological or acidic pH and shown to modify their activity against bacteria and interactions with lipid membranes. In general, AMPs exhibited a lower minimum inhibitory concentration against bacteria at pH 5 than at physiological pH. In vitro assays showed that peptides with charged histidines tend to be less structured (mainly a decrease in helicity) but disrupted lipid membrane integrity more based on dye-leakage assays. The pH-modulated charge state of AMPs can also have an impact on the charge state of the bacterial surface (11).

In this work, we investigated the impact of pH on the histidine-rich AMP caerin 1.1 (Cae-1). Cae-1 is secreted on the skin of the Australian tree frog Litoria (12). It is an amphipathic peptide 25 amino acids in length containing 3 histidines (GLLSVLGSVAKHVLPHVVPVIAEHL-NH₂). Cae-1 adopts a kinked helical structure in the presence of lipid membranes with a hinge induced by two prolines (Pro¹⁵, Pro¹⁹) (13). Two histidines (His¹², His¹⁶) are located on the hydrophilic side of the amphipathic helix (14), while the terminal histidine (His²⁴) is on the hydrophobic face (Fig. 1). The effect of the pH on the peptide interactions with neutral PC or negatively charged PC/PG lipid membranes was investigated by using solid-state NMR, neutron reflectometry, and molecular dynamics simulations.

Figure 1.

Helical wheel (left panel) display of Cae-1 with hydrophobic (yellow), polar (gray), basic (blue), acidic (red), proline (green), and histidine (light blue) residues. The hydrophobic moment is displayed with an arrow. NMR structure (right panel) of Cae-1 with color-coded hydrophobicity surface (blue: hydrophilic, white: neutral, and red: hydrophobic) according to the Kyte and Doolittle scale (15). To see this figure in color, go online.

Materials and methods

Materials

1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC; M_W 678), perdeuterated-acyl chain 1,2-dimyristoyl-sn-glycero-3-phosphocholine (D₅₄-DMPC; M_W 732), 1,2-dimyristoyl-sn-glycero-3-phosphoglycerol (DMPG; MW 688). and perdeuterated-acyl chain 1,2-dimyristoyl-sn-glycero-3-phosphoglycerol (D₅₄-DMPG; M_W 742) were purchased from Avanti Polar Lipids (Alabaster, AL, USA). Cae-1 (GLLSVLGSVAKHVLPHVVPVIAEHL-NH₂; M_W 2584) was synthesized manually by solid-phase peptide synthesis and purified by high-performance liquid chromatography to a purity >95% (Fig. S1) using HCl instead of trifluoroacetic acid salt (16).

Solid-state NMR sample preparation

NMR samples to investigate the interactions of Cae-1 were prepared as follows.

For DMPC-only multilamellar vesicles (MLVs), DMPC was solubilized in chloroform/methanol (3:1 v/v) before removal of solvents under nitrogen flow and further dried under vacuum for 1 h. The lipid film was then hydrated with either a buffer solution at pH 7 (HEPES 20 mM, NaCl 100 mM) or pH 5 (acetate 20 mM, NaCl 100 mM) containing the appropriate amount of Cae-1 to reach a lipid/peptide (L/P) molar ratio of 30:1, except for the peptide-free sample, and lyophilized overnight. The powders were then rehydrated with similar buffer without salt to reach a hydration level of 70 wt %. Three freeze-thaw cycles were used to achieve sample homogeneity prior to packing into a 4 mm rotor (Bruker, Wissembourg, France).

For DMPC/DMPG MLV preparation, DMPC and acyl chain deuterated dimyristoyl-PG D₅₄-DMPG at a 4:1 M ratio were cosolubilized in chloroform/methanol (3:1 v/v) and further processed as for DMPC MLVs.

Solid-state NMR experiments

NMR experiments were conducted on a 9.4 T wide-bore Bruker Avance-III HD NMR spectrometer equipped with a 4 mm triple-resonance magic angle spinning (MAS) probe. The temperature of all experiments was maintained at 303 K.

Static ³¹P NMR spectra were collected under 57 kHz SPINAL64 proton decoupling (17) using a Hahn spin-echo sequence (18), a π/2-pulse with a duration of 4.4 μs, and an echo delay of 30 μs. A minimum of 2000 transients were averaged at a spectral width of 125 kHz and were Fourier transformed following zero filling to 16,000 points and 75 Hz exponential line broadening. ³¹P relaxation experiments were carried out under MAS at 9090 Hz. T₁ relaxation times were measured using the inversion recovery pulse sequence (19). Typical recycle delays were 6 s with variable τ-delay values between 0 and 6 s. T₂ relaxation times were measured with a rotor-synchronized Hahn spin-echo experiment with total echo delay (τ) values between 0.5 and 26 ms using integer multiples of the rotor period. Static ³¹P NMR spectra were analyzed using Topspin 3.5 solids program. Relaxation times were calculated using nonlinear curve fits of peaks integrals versus τ.

Static ²H experiments were performed using a composite pulse solid-echo sequence (20). Operating conditions included 5.5 μs π/2-pulses, 40 μs echo delay, and 0.5 s recycle time. A minimum of 32,000 transients were averaged at a spectral width of 500 kHz and were Fourier transformed following zero filling to 8000 points and 100 Hz exponential line broadening. Static deuterium spectra were numerically de-Paked within NMRPipe (21) using the de-Paking macro. The de-Paked spectra were converted to order parameters by dividing the static splittings with the static coupling constant of 255 kHz (22).

Molecular dynamics (MD)

The starting conformation of Cae-1 was generated from NMR data (23). The CHARMM-GUI membrane builder (24,25) was then used to prepare the DMPC or DMPC/DMPG (4:1) bilayer systems with 200 molecules per leaflet within a rectangular box containing 100 mM KCl salt and a 12.5 Å layer of water. Cae-1 peptides were distributed randomly in a transmembrane orientation with an L/P molar ratio of 30:1, corresponding to 200 DMPC lipids per leaflet and 13 Cae-1 peptides. Of the 13 Cae-1 peptides, two were oriented at 180° compared with the other 11 peptides (Fig. S4). The Charmm36 force field was used (26). Histidine residues were either singly or doubly protonated to model the ionization state expected at pH 5 or 7, and the peptide C-terminus was amidated to match the experimental conditions. This led to Cae-1 having a net charge of +1 at pH 7 or +4 at pH 5. The minimization, equilibration and production runs were performed with the AMBER CUDA package (27) on a desktop machine fitting a GPU GeForce GTX 1080 titanium and a CPU with 12 cores.

Each system was first minimized for 1000 steps using the steepest descent method followed by 1000 steps of the conjugate gradient method with a 12 Å nonbonded interaction cutoff. The peptides and lipids were restrained with a 10 and 2.5 kcal mol⁻¹.Ų potential, respectively. Then, 1.6 ns equilibration MD simulations were run at 310 K, using decreasing positional restraints to maintain the peptide backbone and the lipid atom positions. All covalent bonds involving hydrogen atoms were constrained using the SHAKE algorithm (28), and the rigid internal geometry for TIP3P water molecules was constrained with the SETTLE algorithm (29). The system temperature was maintained at 303 K using a Langevin thermostat (30) with a 3 ps-1 collision frequency. The system pressure was controlled at 1 bar using a semiisotropic Monte-Carlo barostat with a xy surface tension (0 dyne/cm). Fifty production runs of 10 ns each were performed, with each restart using the previous coordinates with random velocities. For all analysis, the first 100 ns of simulation were omitted; thus, the trajectories were analyzed using 400 frames at 1 ns intervals. The lipid order profiles, water contacts, and electron density profiles were thus calculated as an average of 40 simulations each of 10 ns (40 × 10 ns). The area per lipid and peptide tilt angle are presented as a continuous simulation, and averages were calculated across the 40 × 10 ns simulations. For the area per lipid calculation, the peptide cross-section area was removed, using an averaged value of 125 Å for a helical peptide (31).

The MD trajectories were visualized and analyzed using VMD (32) with custom scripts and the CPPTRAJ (33) software, and fitting procedures and plots were created in Gnuplot (http://sourceforge.net/projects/gnuplot).

Results

Interaction of Cae-1 with DMPC MLVs at pH 5 and 7

The impact of Cae-1 on the headgroup region of DMPC MLVs was more significant at pH 5. Cae-1 had no effect on the ³¹P isotropic chemical shift of DMPC but slightly reduced the ³¹P chemical shift anisotropy (using the Haeberlen tensor definition), especially at pH 5 (Fig. 2, A and B; Table 1). At an L/P ratio of 30:1, no isotropic signal was observed in the static ³¹P spectra, indicating that Cae-1 did not induce domains with rapid lipid reorientation, such as toroidal pore or micellar structures, as often observed for AMPs (34,35). However, since the formation of such structures is concentration dependent, the chosen L/P ratio may not be low enough to disrupt significantly the bilayer organization (36). The dynamics of the lipid motions were monitored by measuring the ³¹P relaxation rates under MAS (Fig. 2, C and D; Table 1) (35,37). Cae-1 did not significantly change the spin-lattice T₁ relaxation rate of DMPC at pH 7 but did increase it by ca. 15% at pH 5, which indicates a decrease in intensity of motions on the ns timescale, such as reduced lipid rotation along the lipid long axis. The presence of the peptide reduced the spin-spin T₂ relaxation rate by 32% at pH 5 and 18% at pH 7. T₂ is sensitive to motion on the μs-ms timescale, such as membrane wobbling or lipid diffusion, thus Cae-1 slows global lipid motions. Overall, ³¹P NMR experiments showed that Cae-1 had a stronger impact on the headgroup of DMPC bilayers when the histidine residues are charged, with a decrease in fast and slow motions.

Figure 2.

(A) Static and (B) MAS ³¹P NMR spectra of D₅₄-DMPC MLVs (black lines) at pH 7 (top spectra) and 5 (bottom spectra) and in the presence of Cae-1 (red lines). (C) T₁ spin-lattice relaxation and (D) T₂ spin-spin relaxation experiments with fitted lines (see materials and methods) obtained at 9 kHz. All experiments were performed at an L/P ratio of 30:1 and at 30°C. To see this figure in color, go online.

Table 1.

³¹P NMR parameters of DMPC and DMPC/DMPG (4:1) MLVs upon interactions with Cae-1

	pH 5				pH 7
	CS (ppm)	CSA (ppm)	T1 (s)	T2 (ms)	CS (ppm)	CSA (ppm)	T1 (s)	T2 (ms)
DMPC	−0.88	28.8	0.57 ± 0.01	11.71 ± 0.43	−0.90	26.9	0.84 ± 0.01	16.37 ± 0.58
+ Cae-1	−0.89	27.3	0.65 ± 0.01	8.01 ± 0.17	−0.89	26.8	0.83 ± 0.01	13.46 ± 0.50
DMPC/DMPGa	−0.80	25.7b	0.71 ± 0.01	7.60 ± 0.06	−0.85	25.1b	0.73 ± 0.01	7.21 ± 0.08
+ Cae-1	−0.78	24.1b	0.78 ± 0.01	6.25 ± 0.07	−0.84	24.1b	0.72 ± 0.01	5.96 ± 0.04
DMPC/DMPGa	0.26	25.7b	0.52 ± 0.01	5.84 ± 0.07	0.21	25.1b	0.56 ± 0.03	5.96 ± 0.17
+ Cae-1	0.32	24.1b	0.67 ± 0.01	5.82 ± 0.09	0.25	24.1b	0.60 ± 0.01	5.98 ± 0.06

³¹P NMR parameters from experiments performed at 30°C. SD values were obtained from curve fitting. Cae-1 was added to MLVs at an L/P ratio of 30:1. CS, chemical shift; CSA, chemical shift anisotropy.

^a31P NMR values are for the underlined lipid.

^bA single CSA was fitted for DMPC/DMPG and with Cae-1, as deconvolution of DMPC and DMPG was not possible. Value given according to the Haeberlen tensor definition.

The impact of the peptide on the hydrophobic core of the bilayers was monitored using deuterated DMPC acyl chains via ²H static NMR experiments. The de-Paked spectra were used to obtain the order parameter S_CD profile along the lipid acyl chain (Fig. 3) (38). Cae-1 uniformly increased the order of the DMPC acyl chains at both pH, which suggests that Cae-1 inserted across the DMPC bilayers regardless of the charge of the histidine residues. Furthermore, the ²H spectra of D₅₄-DMPC acquired in the presence of Cae-1 displayed a single set of quadrupolar splittings, indicating that Cae-1 diffused sufficiently quickly to perturb homogeneously the bilayer hydrophobic core without forming more rigid/ordered domains. Ordering of the acyl chains was also observed via neutron reflectometry (NR) for DMPC-supported bilayers at pH 7 (Figs. S2 and S3). NR (supporting material) supported the insertion of the peptide into the hydrophobic core of the bilayer together with some water penetration into the interface also observed.

Figure 3.

(A) De-Paked ²H NMR spectra of D₅₄-DMPC MLVs (black lines) at pH 7 (top spectra) and 5 (bottom spectra) and in the presence of Cae-1 (red lines), and (B and C) the corresponding lipid order profiles. All experiments were performed at an L/P ratio 30:1 and at 30°C. To see this figure in color, go online.

The MD simulations were performed for 40 × 10 ns with Cae-1 peptides initially positioned in a transmembrane orientation at random positions and with an L/P molar ratio of 30:1, which translated to 200 DMPC lipids per leaflet and 13 Cae-1 peptides. Of the 13 Cae-1 peptides, two were oriented at 180° compared with the other 11 peptides in order to test if specific peptide-peptide interactions are more likely via parallel or antiparallel assembly (see Fig. S4). The location of the peptides was chosen by taking into consideration the NMR and NR experimental data, suggesting deep penetration of Cae-1 into the lipid bilayers. The DMPC bilayer was simulated once, and the pH 7 and 5 simulations were run with neutral (i.e., only N_D1 is protonated, net charge +1) or charged (i.e., both imidazole nitrogens are protonated, net charge +4) histidine residues, respectively. From the simulations, the peptide tilt angle lengthways of their long axis versus the bilayer normal; the electron density profile including those of the three histidine residues and their hydration shell; the lipid order profiles; and the areas per lipid were extracted as an average of the 40 × 10 ns simulations (Figs. 4 and S5). In the presence of Cae-1, water penetrated deeply into the headgroup of the bilayer with greater effect at pH 5. His¹² and His²⁴ remained at similar positions at pH 7 and 5, but in the latter case, the residues spanned a larger area. The area per lipid was slightly increased in the presence of Cae-1 at pH 5 but was similar at pH 7 compared with the peptide-free DMPC bilayer. His¹⁶ exhibited a more central location at pH 7 but was distributed across two populations further away from the center of the bilayer at pH 5. The number of water molecules surrounding the histidine residues (within 5 Å) was constant but significantly higher at pH 5, especially for His¹² and His¹⁶ (Fig. S5). It is also worth noting that at pH 7, not all peptide histidine residues were near a phosphate group, but almost all were at pH 5 (Fig. S5). Furthermore, the charged Lys¹¹ and Glu²³ residues were slightly more hydrated at pH 5 with the cationic residue in proximity to ³¹P atoms. Interestingly, the peptide tilt angles were not very dispersed, centered at ca. 35°, at pH 7 but displayed a large distribution at pH 5, centered at ca. 42°. (Fig. S6). Overall, the MD simulations showed similar effects on the DMPC bilayer dynamics and packing properties analogous to the NMR and NR (see Table S2) experimental data. MD simulations allowed investigation at greater atomic detail of the location of the peptides within the bilayer, which also showed stronger perturbation with the charged histidine residues. In particular, His¹² and His¹⁶ seemed to displace water toward the center and likely shifted the transmembrane orientation of Cae-1. However, since the MD simulations were not run for a very long time, it is unlikely that the systems had reached final equilibrium, and the observed perturbations may be transient, although they do support the experimental data.

Figure 4.

Electron density profile obtained from 40 × 10 ns simulations along the z axis (normal) of (A) the lipid bilayer and in the presence of Cae-1 at (B) pH 7 and (C) 5. Phosphorous atoms (orange circles), water (cyan triangles), His¹² (magenta crosses), His¹⁶ (pink squares), and His²⁴ (purple inverted triangles) are depicted. (D) Order parameter profiles and (E) area per lipid obtained from 40 × 10 ns simulations of the DMPC bilayer (black triangles) and in the presence of Cae-1 at pH 7 (red squares) and 5 (blue circles). MD simulations were run at 303 K. To see this figure in color, go online.

Interaction of Cae-1 with DMPC/DMPG (4:1) MLVs at pH 5 and 7

Bacterial lipid membranes are mostly negatively charged due to the high content of PG and, to a much lesser amount, cardiolipin phospholipids (3). The impact of PG lipids on the interactions of Cae-1 at pH 5 and 7 was thus investigated similarly to the DMPC systems. The static ³¹P NMR experiments did not exhibit any significant isotropic contribution, but the presence of Cae-1 slightly reduced the chemical shift anisotropy regardless of the pH. The MAS ³¹P spectra displayed small upfield shifts of ca. 0.05 ppm for DMPG but not so for DMPC (Fig. 5; Table 1). Relaxation rates were perturbed similarly as for DMPC systems. T₁ values were longer in the presence of the peptide, especially for DMPG at pH 5, which exhibited the greatest change of ca. 29% compared with 10% for DMPC. T₂ values were shorter for DMPC but did not significantly change for DMPG. Thus, the overall effect of Cae-1 on the lipid headgroup of DMPC/DMPG bilayers was to reduce the slow and fast motions of the phosphate group, with more effect on the DMPG headgroup and when the histidine residues are charged. Thus electrostatic interactions appear to have higher impact on the charged membrane surface, as previously reported (39).

Figure 5.

(A) Static and (B) MAS ³¹P NMR spectra of DMPC/D₅₄-DMPG (4:1) MLVs (black lines) at pH 7 (top spectra) and 5 (bottom spectra) and in the presence of Cae-1 (red lines). T₁ spin-lattice relaxation experiments of (C) DMPC and (D) D₅₄-DMPG. T₂ spin-spin relaxation experiments of (E) DMPC and (F) D₅₄-DMPG with fitted lines (see materials and methods) obtained at 9 kHz MAS. All experiments were performed at an L/P ratio of 30:1 and at 30°C. To see this figure in color, go online.

The perturbation of the DMPC/DMPG (4:1) MLV hydrophobic core upon the addition of Cae-1 was monitored using D₅₄-DMPG (Fig. 6). Deuteration of DMPG was chosen since stronger interactions between the negatively charged lipid and the positively charged peptide were anticipated, and ²H signals would not be distinguishable if both lipids were deuterated. As for the D₅₄-DMPC systems, Cae-1 produced a significant increase in the order of the D₅₄-DMPG acyl chains. The ordering of the anionic lipid was sensed along the entire chain, thus indicating that Cae-1 remained inserted across the lipid bilayer also in the charged MLVs. Again, a single set of quadrupolar splittings was observed, signifying that all D₅₄-DMPG lipids were in a similar environment. At an L/P ratio of 30:1 and a DMPC/DMPG molar ratio of 4:1, the DMPG/Cae-1 molar ratio is 6:1. Since the ³¹P NMR data did not indicate much segregation of the PG lipids by Cae-1, as the chemical shift and relaxation rates of DMPC in the PC/PG bilayers did not tend toward those of the pure DMPC system, Cae-1 is also likely diffusing sufficiently quickly and homogenously perturbing the DMPC and D₅₄-DMPG lipids.

Figure 6.

(A) De-Paked ²H NMR spectra of DMPC/D₅₄-DMPG (4:1) MLVs (black lines) at pH 7 (top spectra) and 5 (bottom spectra) and in the presence of Cae-1 (red lines), and (B and C) corresponding lipid order profiles. All experiments were performed at an L/P ratio of 30:1 and at 30°C. To see this figure in color, go online.

The MD simulations were performed as for the DMPC systems. The starting orientations of Cae-1 were identical, only the lipid systems changed with 160 DMPC and 40 DMPG used per leaflet. Overall, similar observations were obtained: the lipid headgroups tended to be more dispersed along the bilayer normal axis, moving inward, and water penetrated deeper into the headgroup area (Figs. 7 and S7). The perturbation of the acyl chains of DMPC and DMPG by Cae-1 at both pH 5 and 7 were less pronounced than for the DMPC bilayer and differed depending on the charge of the peptide. At pH 7, similar increases in the order parameter of DMPC and DMPG were observed, but at pH 5, the order parameter of C1 to C6 near the glycerol region exhibited little perturbation for both DMPC and DMPG, while the rest of the chain carbons were ordered by the peptide, as was observed by ²H NMR. The area per lipid increased in the presence of Cae-1 at pH 5 but not at pH 7 (Fig. 7), as observed with the DMPC bilayer. The hydration of the histidine residues and proximity of a phosphate group within 5 Å of the residue were also not much different from that found in the DMPC systems, regardless of the charge of Cae-1 (Fig. S7). Interestingly, the peptide tilt angles were more dispersed at pH 7 than was observed for the DMPC membrane system, while at pH 5, the same larger dispersion was measured (Fig. S8).

Figure 7.

Electron density profile obtained from 40 × 10 ns simulations along the z axis (normal) of (A) the DMPC/DMPG (4:1) lipid bilayer and in the presence of Cae-1 at (B) pH 7 and (C) 5. PC phosphorous atoms (orange circles), PG phosphorous atoms (red squares), water (cyan triangles), His¹² (magenta crosses), His¹⁶ (pink squares), and His²⁴ (purple inverted triangles) are depicted. Lipid order parameter profiles for (D) DMPC and (E) DMPG, and (F) area per lipid of the lipid bilayer (black triangles) and in the presence of Cae-1 at pH 7 (red squares) and 5 (blue circles) obtained from 40 × 10 ns simulations. MD simulations were run at 303 K. To see this figure in color, go online.

Discussion

AMPs are membrane-active peptides, their primary target is the lipid membrane, and they have a greater preference for prokaryotic rather than eukaryotic membranes. However, it has been shown that their disruption of neutral membranes is greater despite a higher affinity for negatively charged membranes (40). Thus, electrostatic interactions are the driving force for AMP attraction toward cell membranes. Since several amino acids display an ionization state that is pH dependant, the impact of pH on AMP activity is of importance for the goal of using AMPs as antibiotic alternatives (41). Membrane-permeabilizing peptides have been designed that become active at low pH and are engineered to serve as intracellular carriers (42). Interestingly, histidine residues are commonly found in AMPs with a similar frequency as for mammalian proteins (2.1% vs. 2.3%, respectively). Furthermore, over 35% of AMPs in the APD3 AMP database display at least one histidine residue, with >15% having several (43), and Cae-1 having three histidines.

In the present study, Cae-1 did not drastically change its activity based on the charge of the lipid membrane, which was previously studied at pH 7 in DMPC and DMPC/DMPG (2:1) by surface plasmon resonance and solid-state NMR (39). However, both membrane systems were more disturbed by Cae-1 at pH 5 than pH 7. The AMP primarily slowed the dynamics of the headgroup and hydrophobic core of the phospholipid membranes, likely as a result of inserting within the bilayers. Hence, it is unlikely that Cae-1 adopted a different orientation parallel or perpendicular to the bilayer normal, based on the pH as reported for other membrane-active peptides, such as LAH4 (44) or GWALP23 (45). This is likely due to the high content of hydrophobic residues and the amphipathic structure of the helical Cae-1 (14,46). However, the MD simulations showed that, although maintaining an overall transmembrane orientation, the tilt angle fluctuation of the peptide long axis was significantly more distributed at pH 5 than pH 7. This was probably due to the greater exposure of the charged histidine residues to water, similarly to the snorkeling-out effect seen with lysine and arginine residues (47,48,49). Furthermore, pH can also modify the ionization state and surface potential of the lipid membranes, which could also play a role in modulating the interactions with the cationic peptides (50,51). As is often the case for amphipathic AMPs, electrostatic and hydrophobic interactions are both playing a critical role in the peptide-lipid macromolecular assembly.

Note that in this study, the L/P ratio was fixed at 30:1. It is possible that the chosen ratio is not representative of the local peptide concentration that would be necessary to induce a major disruption of the lipid bilayers. Increasing the peptide concentration is likely a prerequisite to trigger particular peptide-peptide assembly, such as transmembrane pores (52), as observed for pore-forming proteins (53). However, even below the minimum inhibition concentration for the AMP maculatin 1.1, transient pore formation in phospholipid bilayers and secondary targets in bacteria have been observed (54,55).

Conclusions

Although further investigations are necessary to fully characterize the pH dependence of Cae-1 in regard to its antimicrobial activity, the peptide does appear to act as transmembrane peptide in zwitterionic and anionic bilayers, and its effect is modulated by pH. Not only are the peptide concentration and charge state important parameters in peptide-peptide self-assembly, the lipid composition also plays a critical role for the interactions. As well as further studies in model membrane systems to characterize the mode of action, recent developments in in-cell NMR allow studies of AMPs in bacteria (54,55,56,57,58) or even proteins (59,60,61,62) within live cells.

Author contributions

M.-A.S. and F.S. designed the research; M.-A.S., A.P.L.B., T.A., and S.R. performed the research and analyzed the data; M.-A.S., A.P.L.B., S.R., and F.S. wrote the manuscript; and M.-A.S. and F.S. edited the manuscript.

Editor: Richard Pastor.