Melittin-Induced Lipid Extraction Modulated by the Methylation Level of Phosphatidylcholine Headgroups

Abstract

Protein- and peptide-induced lipid extraction from membranes is a critical process for many biological events, including reverse cholesterol transport and sperm capacitation. In this work, we examine whether such processes could display specificity for some lipid species. Melittin, the main component of dry bee venom, was used as a model amphipathic α-helical peptide. We specifically determined the modulation of melittin-induced lipid extraction from membranes by the change of the methylation level of phospholipid headgroups. Phosphatidylcholine (PC) bilayers were demethylated either by substitution with phosphatidylethanolamine (PE) or chemically by using mono- and dimethylated PE. It is shown that demethylation reduces the association of melittin with membranes, likely because of the resulting tighter chain packing of the phospholipids, which reduces the capacity of the membranes to accommodate inserted melittin. This reduced binding of the peptide is accompanied by an inhibition of the lipid extraction caused by melittin. We demonstrate that melittin selectively extracts PC from PC/PE membranes. This selectivity is proposed to be a consequence of a PE depletion in the surroundings of bound melittin to minimize disruption of the interphospholipid interactions. The resulting PC-enriched vicinity of melittin would be responsible for the observed formation of PC-enriched lipid/peptide particles resulting from the lipid efflux. These findings reveal that modulating the methylation level of phospholipid headgroups is a simple way to control the specificity of lipid extraction from membranes by peptides/proteins and thereby modulate the lipid composition of the membranes.

Introduction

Protein- and peptide-induced lipid extraction from membranes is a critical process for many biological events. For example, binder-of-sperm protein 1 (BSP1), the most abundant protein of bovine seminal plasma, extracts lipids (more specifically, phosphatidylcholines (PCs) and cholesterol (Chol)) from sperm membranes, and this phenomenon is an essential step in sperm capacitation (1, 2, 3). Apolipoproteins, such as ApoA1, are another species involved in lipid extraction. They extract Chol and phospholipids from peripheral tissues to form high-density lipoproteins that are carried toward the liver for catabolism (4). This phenomenon is pivotal in reverse Chol transport and consequently is intimately related to the risk of coronary heart diseases. The mechanism by which lipids are extracted from cells encompasses a complex series of reactions involving several molecular species (5), but at the end of the process, suitable lipid selectivity must be achieved. These examples illustrate the complexity of membrane lipid extraction processes. In spite of their importance, however, the general molecular interactions that govern their underlying mechanism and, more specifically, the basis of their lipid selectivity are still largely unknown.

In this work, we sought to gain insights into membrane lipid extraction processes by studying the lipid efflux induced by melittin from model membranes. Melittin, the main component of dry bee venom (6), is a 26-amino-acid peptide that has been widely used as a model peptide for various purposes over the last decades (for a general review, see Raghuraman and Chattopadhyay (7)). Like ApoA1, melittin acts as a helical amphiphile that can solubilize lipids. This property is due to its amphipathic character coupled with a highly charged C-terminal region (Lys-Arg-Lys-Arg). Melittin also displays an appealing therapeutic potential for diverse applications, and various reviews have discussed its antimicrobial properties (8, 9), its potential for cancer therapy (10, 11), treatment (12) and prevention of HIV (13), and treatment of parasitical infections (14), to name a few.

Melittin interacts spontaneously with phospholipid membranes. At low concentrations, it binds to membranes as an amphipathic α-helix (15, 16, 17), induces pore formation, and causes leakage (18, 19, 20, 21, 22, 23). At higher concentrations, melittin leads to lipid extraction by forming soluble small lipid/peptide bicelles or nanodiscs (24, 25, 26). Melittin-induced lipid extraction was shown to be modulated by the addition of Chol (27, 28), phosphatidylethanolamine (PE) (29), or negatively (15, 30, 31) or positively (31) charged phospholipids in membranes. Although these non-PC lipids have been reported to inhibit the formation of melittin/lipid bicelles, the cause of this inhibition is different for different lipid species. On one hand, it was reported that Chol, PE, and positively charged phospholipids simply prevent the association of melittin with bilayers. On the other hand, it was shown that negatively charged phospholipids enhance the affinity of melittin for bilayers but anchor the peptide at the bilayer interface, preventing the peptide relocation that is necessary to induce lipid extraction. It is solidly established that melittin-induced lipid extraction is modulated by the bilayer composition, but very little is known about the lipid specificity of the extraction. A previous study showed that particles resulting from melittin-induced extraction from 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DPPC)/Chol bilayers were depleted in Chol compared with the sterol content of the original bilayers (27). This result suggests that melittin does not merely extract membrane lipids as a universal detergent.

Our main goal in this work was to determine the effect of PE headgroups on the lipid extraction induced by melittin and, more specifically, on its lipid selectivity. Elucidating the impact of methylation of the headgroup (by comparing PE and PC) on peptide-induced lipid extraction would aid in defining the molecular mechanism of this process. It was previously shown that melittin has a reduced affinity for pure 1,2-dielaidoyl-sn-glycero-3-phosphoethanolamine (DEPE) bilayers (29). Moreover, the negative curvature of PE was proposed to inhibit the formation of positively curved toroidal pores by melittin, and therefore to limit peptide-induced leakage (32, 33, 34, 35, 36). To examine the influence of the methylation level of the phospholipid headgroup on melittin-induced lipid extraction, we characterized melittin binding to model membranes and the resulting lipid efflux. Model membranes were prepared from DPPC/1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine (DPPE), and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC)/1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) mixtures in various proportions, as well as from mono- and dimethyl DPPE (1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-methyl (DPMePE) and 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N,N-dimethyl (DPMe₂PE), respectively). These two series of measurements provided two different ways to modulate the methylation level of the ammonium groups at the bilayer interface. Fluorescence measurements of the tryptophan at position 19 of melittin were conducted to investigate melittin binding to bilayers. We analyzed the extent of lipid extraction by melittin using a procedure analogous to one reported previously (37). We examined the composition of the lipid/peptide particles that resulted from the extraction, as well as that of the remaining bilayers, using liquid chromatography coupled with mass spectrometry (LC-MS) to determine any selectivity toward one lipid or another.

Materials and Methods

Chemicals

Melittin was purified from bee venom (Sigma, St. Louis, MO) by ion exchange chromatography on SP-Sephadex C-25 (38). POPC, POPE, DPPC, DPPE, DPMe₂PE, and DPMePE were purchased from Avanti Polar Lipids (Alabaster, AL). EDTA, NaCl, and 3-[N-morpholino]propanesulfonic acid (MOPS) were obtained from Sigma. All chemicals were used as received.

Lipid membrane preparation

Individual lipids were dissolved in a benzene/methanol mixture (90:10 (v/v)), mixed to obtain the desired molar ratio, and then lyophilized. The lipid powders were hydrated in a MOPS buffer (50 mM) containing 100 mM NaCl and 100 μM EDTA, pH 7.4. The samples were submitted to three freeze-and-thaw cycles (from the liquid nitrogen temperature to 65°C) to form multilamellar vesicles (MLVs) used for the extraction experiments. For the fluorescence experiments, the MLVs were extruded at 75°C using a commercial extruder (Northern Lipids, Vancouver, Canada) to obtain 100-nm-diameter large unilamellar vesicles (LUVs).

Extraction

Melittin was dissolved in the MOPS buffer and its concentration was determined by its absorbance at 280 nm, using a molar absorptivity coefficient of 5570 M⁻¹ cm⁻¹ (39). The lipids and the peptide were mixed in microcentrifuge tubes to obtain the desired molar lipid-to-melittin incubation ratio (L/M) and a fixed phospholipid concentration of 1 mg/mL. The suspensions were then incubated at 70°C for at least 30 min. After the incubation, the samples were centrifuged for 5 min at 20,800 g and 1°C. The centrifugation of control samples (without melittin) showed that more than 95% of lipids were found in the pellets for all mixtures. The supernatants were isolated and the pellets were resuspended in the MOPS buffer for analysis. The experiments were carried out in triplicates unless stated otherwise.

Lipid analysis

We determined the phospholipid content in the supernatants and pellets using Bartlett’s phosphorus assay (40). The extent of extraction was calculated as

$$\mathit{Extraction}\%=\mathit{phospholipids}\mathit{in}\mathit{supernatant}/\mathit{total}\mathit{phospholipids}.$$ (1)

We performed an LC-MS analysis to determine the lipid composition of the supernatants and pellets, using an Agilent Technologies 1100 series system equipped with an 1100 MSD mass spectrometer. Samples were eluted on a YMC diol column (4.6 × 150 mm, 5 μm particle size; Agilent Technologies) maintained at 50°C. Elution of the phospholipids was achieved over 10 min, using an acetonitrile/water (100 mM ammonium acetate) 91/9 mixture at a 0.6-ml/min flow rate. An electrospray ionization source was used in the positive-ionization mode. Nitrogen was used as the drying gas at 250°C and 12 L/min. The nebulizing gas was also nitrogen, held at 241 kPa. The analysis was conducted in the single-ion-monitoring mode with a dwell time of 290 ms. We determined the extent of extraction for each lipid species using the following equation:

$$Extraction\%=\frac{A_s}{A_s+A_p},$$ (2)

where A_s and A_p are the lipid peak areas from the supernatant and pellet analyses, respectively.

Fluorescence measurements

We recorded fluorescence spectra using a Photon Technology International fluorometer with bandwidths of 1.0 and 2.0 nm for the excitation and emission monochromators, respectively. The excitation wavelength was set at 283 nm. To examine melittin association with lipid bilayers as a function of temperature, we first mixed melittin (0.9 μM) with LUVs (L/M = 400) in a sample cell kept at 5°C. The fluorescence spectra were then acquired at different temperatures ranging from 5°C to 90°C, and then back to 5°C. For each temperature, a 10-min equilibration period was introduced before data acquisition. For the binding studies, melittin was first added to a cell, and aliquots of a lipid suspension were added to vary L/M from 0 to 400. After each lipid addition, the fluorescence spectrum of melittin was acquired at 65°C and then at 20°C. An equilibration period of at least 20 min after the desired temperature was reached was introduced before data were acquired. For analysis of the supernatants and pellets that resulted from the extraction experiments, the samples were directly transferred to a cell and measured at 20°C (the pellets were previously resuspended in MOPS buffer).

Results

Association with LUVs

The hypsochromic shift of tryptophan fluorescence that is observed when this amino acid is transferred from water to a more hydrophobic environment is a useful tool for characterizing the peptide association with membranes (41). We characterized the binding of melittin to membranes made of DPPC/DPPE as a function of temperature by plotting the evolution of the tryptophan fluorescence band maximum (λ_max; Fig. 1).

Figure 1.

Evolution of tryptophan fluorescence as a function of temperature for melittin in the presence (L/M = 400) of DPPE (dark green arrows), DPPC/DPPE 25/75 (light green arrows), DPPC/DPPE 50/50 (orange arrows), DPPC/DPPE 75/25 (maroon arrows), or DPPC (red arrows) LUVs, from 5°C to 90°C and then back to 5°C. ▴ was obtained for melittin in solution. To see this figure in color, go online.

The λ_max of melittin in the presence of DPPC or DPPC/DPPE 75/25 LUVs, at 5°C, was observed at 337 and 343 nm, respectively. This hypsochromic shift, compared with the 349 nm obtained for free melittin, suggested a lipid-melittin association with these gel-phase membranes. No association was observed with DPPC/DPPE 50/50, DPPC/DPPE 25/75, or DPPE membranes in the gel phase, as λ_max was observed at 349 nm.

Heating of the membranes resulted in a hypsochromic shift for all of the DPPC-containing mixtures, indicating an insertion or a deeper insertion of the peptide into the bilayer. For DPPC, this shift occurred between 25°C and 35°C, with λ_max decreasing to 331 nm, a value that is typical for completely membrane-bound melittin. It should be noted that 35°C corresponds to the pretransition temperature of DPPC bilayers. It has been shown that melittin added to gel-phase DPPC penetrates the membranes and forms bicelles at the pretransition temperature (27, 42). An analogous fluorescence shift was observed for the DPPC/DPPE mixtures. The temperature at which the shift was observed depended on the composition: it was between 30°C and 50°C for DPPC/DPPE 75/25 and 50/50 mixtures, and between 45°C and 70°C for DPPC/DPPE 25/75 bilayers. Thus, for these three mixtures, the hypsochromic shift progressed over a larger temperature span than was observed for DPPC, beginning a few degrees below the apparition of the P_β or L_α phase and ending approximately at a temperature where the lipid bilayers were exclusively in the L_α phase, as inferred from the DPPC/DPPE phase diagram (43). The L_β-to-L_α phase transition was observed over temperature ranges of 35–50°C, 42–57°C, and 49–62°C for the DPPC/DPPE 75/25, 50/50, and 25/75 mixtures, respectively. λ_max measured at high temperatures was also sensitive to the bilayer composition, being 333 nm for DPPC/DPPE 75/25 bilayers, 338 nm for the equimolar mixture, and 341 nm for DPPC/DPPE 25/75 bilayers. These results suggested that the DPPE content of the bilayers limits the depth of penetration of melittin and/or limits the proportion of membrane-bound melittin. A fluorescence maximum typical of free melittin was observed at all temperatures in the case of pure DPPE membranes (T_m = 64°C), suggesting essentially no insertion of the peptide into these membranes.

The reversibility of melittin association upon cooling was assessed for the different lipid mixtures. First, λ_max decreased to 328 nm when DPPC membranes were cooled to 5°C, indicating a strong, irreversible association with the gel-phase membrane. For DPPC/DPPE 75/25 bilayers, λ_max increased back to ∼337 nm upon cooling to 20°C. This is an indication that a certain proportion of melittin relocated once the bilayers were cooled to the gel phase, either by dissociating completely from the membrane or, at least, by reorienting the tryptophan toward a more polar environment. For DPPC/DPPE 50/50 and 25/75, the fluorescence shift observed during the lipid phase change was practically fully reversible, with λ_max corresponding to that of free melittin at low temperatures. It is inferred that these mixtures gave rise to a complete dissociation of melittin upon cooling to the gel phase.

To further describe the thermal partial reversibility of melittin association with DPPC/DPPE 75/25 membranes, we characterized melittin binding to these membranes as a function of the lipid/melittin ratio at 20°C and 65°C, and compared this behavior with that of pure DPPC (Fig. 2). A melittin solution was titrated with LUVs. After each addition of LUVs, λ_max was measured at 65°C and then measured again after cooling to 20°C to portray the reversibility of the peptide association. For DPPC bilayers, the shift of λ_max obtained as a function of L/M was very similar in the gel and fluid phases, with the overall shift being slightly larger for fluid DPPC bilayers. These affinity curves were similar to those reported in a previous study (44). Peptide association with the bilayers was maintained in the gel phase. The association curves obtained for melittin with DPPC/DPPE 75/25 bilayers displayed a different behavior. The shift of λ_max observed upon addition of fluid-phase bilayers was very similar to that obtained with pure DPPC bilayers, indicating a similar melittin/bilayer association. However, in contrast to the DPPC behavior, the fluorescence band shifted back to a higher wavelength when the bilayers were cooled to the gel phase. This finding suggests that the addition of 25% DPPE to DPPC membranes had no significant effect on peptide insertion into fluid-phase bilayers, but induced dissociation of the peptide from gel-phase bilayers upon cooling.

Figure 2.

Evolution of tryptophan fluorescence characteristic of melittin association with DPPC (diamonds) or DPPC/DPPE 75/25 (triangles) bilayers during incubation at 65°C (solid symbols) and subsequent cooling to 20°C (open symbols) after each addition of lipids. To see this figure in color, go online.

The variation of melittin λ_max was also determined in the presence of membranes of DPMe₂PE (T_m = 48°C) and DPMePE (T_m = 58°C) (45) (Fig. 3), an alternative approach to vary the methylation level of the headgroups. In addition, melittin association with POPE LUVs (T_m = 25°C) was determined to assess the affinity of melittin for PE membranes that are less ordered than DPPE LUVs. For the three systems, a λ_max typical of free melittin was observed at temperatures where the bilayers were in the gel phase. Upon heating, λ_max shifted to 330–335 nm, indicating association of the peptide with the bilayers. This shift was observed at 20–50°C, 43–50°C, and 55–70°C for POPE, DPMe₂PE, and DPMePE membranes, respectively. It appeared that the peptide insertion was initiated a few degrees before the gel-to-fluid phase transition of the phospholipids. Cooling down the membranes resulted in the recovery of free-melittin λ_max roughly over the same temperature intervals, indicating the full reversibility of melittin association with these membranes.

Figure 3.

Evolution of tryptophan fluorescence as a function of temperature for melittin in the presence (L/M = 400) of DPPE (dark green arrows), DPMePE (purple arrows), DPMe₂PE (blue arrows), POPE (black arrows), or DPPC (red arrows) LUVs from 5°C to 90°C and then back to 5°C. ▴ was obtained for melittin in solution. To see this figure in color, go online.

Lipid extraction by melittin

We sought to determine the influence of the substitution of PC by PE on lipid extraction using DPPC/DPPE and POPC/POPE membranes. The MLVs and melittin (L/M = 20) were first incubated above the membrane transition temperature and then cooled down to 1°C for centrifugation. It is hypothesized that the remaining MLVs (with bound melittin, if present) pellet, whereas the extracted lipids, forming small self-assemblies with melittin, remain in suspension. The total lipid extraction was quantified for mixtures with various PE contents (Fig. 4 A). Melittin extracted 86% of the lipids from pure DPPC membranes. This level is consistent with previous NMR results that showed a similar proportion of lipids involved in an isotropic phase (i.e., bicelles) (31). Melittin-induced lipid extraction with pure POPC bilayers yielded almost the same proportions, as was previously observed with egg-PC using quasi-elastic light scattering (24). The inclusion of PE in PC membranes inhibited melittin-induced lipid extraction. For both DPPC and POPC, the lipid extraction was reduced by ∼50% for membranes with 25% PE compared with those of pure PC. Membranes with more PE offered a greater resistance to melittin-induced extraction, and no lipid could be detected in the supernatant of the membranes with ≥75% PE, for an L/M of 20. This finding is consistent with a previous study indicating that melittin had no impact on the ³¹P NMR spectrum of gel-phase DEPE at L/M = 20 (29). Our results indicate that the inhibitory effect of PE is progressive and is also observed with DPPE as well as POPE.

Figure 4.

(A and B) Lipid extraction (A) and wavelength of fluorescence maximum in supernatants at 20°C (B) after incubation (L/M = 20) of melittin with MLVs of DPPC/DPPE (squares) or POPC/POPE (circles). To see this figure in color, go online.

The state of melittin in the supernatant after incubation with lipid bilayers and centrifugation was characterized by the position of its fluorescence maximum (Fig. 4 B). The obtained values were consistent with the reported lipid extraction proportions (Fig. 4 A). For DPPC and POPC bilayers, the fluorescence maximum appeared between 335 and 341 nm, and these values were shifted relative to that observed for free melittin (350 nm). This hypsochromic shift was associated with the contribution of melittin included in the small lipidic complexes, where its tryptophan should be in an apolar environment. It should be noted that at an L/P of 20, a considerable proportion of free melittin coexisted with its bound form (see Fig. 1). Conversely, the fluorescence band of melittin in the supernatant after incubation with DPPE or POPE bilayers was found at 349 nm, a value that is typical of the peptide free in solution and consistent with the fact that no lipid extraction was observed. The progressive decrease in lipid extraction with an increasing PE proportion in bilayers was effectively reflected by the progressive shift of the melittin fluorescence band in the supernatant.

Melittin-induced lipid extraction was also modulated by methylation of the lipid headgroups (Fig. 5 A). The removal of one methyl from DPPC (DPMe₂PE) reduced the lipid extraction by more than half for an L/M ratio of 20, from 86% with DPPC to 37% with DPMe₂PE bilayers. The removal of a second methyl group on the PC headgroup (DPMePE) almost completely inhibited the lipid extraction, similar to what was observed with DPPE. As in the case of the PC/PE mixtures, the fluorescence maximum of melittin in the supernatant varied in a consistent manner with the lipid extraction level. The position of the band was typical of free melittin in the case of the incubation with DPMePE (which led to no lipid extraction). Melittin fluorescence maximum in the supernatant after incubation with DPMe₂PE bilayers was shifted to 348 nm, indicative of the coexistence of the lipid-bound and free melittin states.

Figure 5.

(A and B) Quantification of lipid extraction (A) and wavelength of fluorescence maximum in the supernatant at 20°C (B) after incubation (L/M = 20) of melittin with MLVs of DPPE, methylated DPPE (DPMePE, DPMe₂PE), or DPPC.

The evolution of the phospholipid extraction as a function of melittin concentration was investigated in more detail for some DPPC/DPPE mixtures (Fig. 6). The capacity of melittin to extract lipids from bilayers was dependent on their lipid composition. The lipid extraction increased abruptly with melittin concentration for DPPC bilayers, from 3% at M/L = 15 × 10⁻³ (L/M = 67) to 86% at M/L = 50 × 10⁻³ (L/M = 20). This result is consistent with the evolution of the area of the narrow line, associated with rapidly tumbling bicelles, as a function of melittin content in ²H-NMR spectra (31). In general, the presence of DPPE in DPPC membranes decreased their susceptibility to melittin-induced lipid extraction: for M/L = 100 × 10⁻³ (L/M = 10), the extent of the lipid extraction decreased from 90% for pure DPPC bilayers to 50% for DPPC/DPPE 75/25 bilayers, and to 30% for DPPC/DPPE 50/50 bilayers. A progressive and monotonous lipid extraction was obtained as a function of increasing peptide amount for DPPC/DPPE 50/50 bilayers. However, the evolution of the lipid efflux as a function of melittin proportion displayed a more complex pattern for DPPC/DPPE 72/25 bilayers. The lipid extraction appeared to be more efficient than that observed with DPPC bilayers for low melittin contents, up to an M/L ratio of 30 × 10⁻³ (L/M = 33). For higher melittin contents, the extent of lipid extracted by melittin seemed to reach a plateau. This complex behavior is discussed below.

Figure 6.

Quantification of lipid extraction after incubation of melittin with MLVs of DPPC (diamonds), DPPC/DPPE 75/25 (triangles), or DPPC/DPPE 50/50 (circles) at different ratios. The incubation ratio (x axis) is displayed as melittin per 1000 lipids (bottom), as well as L/M (top).To see this figure in color, go online.

Lipid selectivity of melittin-induced extraction

We determined the composition of the extracted lipid fraction after incubation of melittin with PC/PE bilayers to identify whether the extraction was specific for a lipid species. The proportions of extracted PC and PE from bilayers made from mixtures of DPPC/DPPE or POPC/POPE are presented in Fig. 7. The results reveal that PC was systematically more extracted than PE from the original bilayers. The extent of melittin-induced extraction was 1.7–2.1 times larger for DPPC than for DPPE in DPPC/DPPE 75/25 bilayers, for M/L between 50 × 10⁻³ and 100 × 10⁻³ (L/M from 20 to 10). The lipid selectivity of the extraction was even more marked for DPPC/DPPE 50/50 membranes, with the proportion of extracted DPPC being three to six times greater than that of extracted DPPE.

Figure 7.

(A and B) Melittin extraction selectivity after incubation with DPPC/DPPE 75/25 (A) or DPPC/DPPE or POPC/POPE 50/50 (B) MLVs. The black bars represent PC extraction and white bars represent PE extraction. The incubation ratio (x axis) is displayed as melittin per 1000 lipids.

Discussion

Effect of demethylation on melittin association

The binding experiments (Figs. 1 and 3) show that the presence of three methyl groups on the ammonium of the phospholipid headgroup is essential for the association of melittin with gel-phase membranes. A reduction of the methylation level, either by the substitution of PC by PE or by the use of demethylated PC, led to a reversible fluorescence shift after cooling to the gel phase that was indicative of the dissociation of melittin from the lipid bilayers. The association of melittin with fluid bilayers was considerably favored compared with the gel-phase ones, as was demonstrated by the temperature-dependent hypsochromic shift of λ_max for the lipid mixtures whose behavior is reported in Figs. 1 and 3. Melittin association with bilayers involves penetration of the peptide into the apolar core of the bilayers, as was initially suggested by an analysis of an x-ray melittin structure that revealed its amphipathic, α-helical character (46). Using the parallax method, previous studies estimated the depth of penetration of the 19-Trp residue for melittin bound to 1,2-di-oleoyl-sn-glycero-3-phosphocholine (DOPC) bilayers to be 10.6 Å from the bilayer center (47, 48). The insertion of melittin into the hydrophobic core of POPC membranes was also demonstrated by the quenching of 19-Trp fluorescence upon addition of brominated lipids (49, 50). Calculations of the solvation free energy of melittin (51) demonstrated that its binding at the membrane interface was driven by strong hydrophobic interactions, which overcame the energy cost of transferring the polar backbone of the peptide from the high-dielectric aqueous phase to the apolar bilayer core.

The strong intermolecular interactions between PE headgroups may explain the reduced interaction between melittin and bilayers with a low interfacial methylation level of phospholipid ammonium groups. PEs are known to form membranes with a tighter phospholipid packing associated with the small van der Waals volume of their headgroups (52) and their capacity to form intermolecular hydrogen bonds (53). These features induce ordering of the lipid chains, as shown by the larger fluid-phase chain order parameters and the derived smaller chain cross-sectional area of POPE compared with POPC (54, 55, 56). The tighter chain packing induced by the decrease of the methylation level of the interfacial ammonium groups is proposed to be thermodynamically unfavorable for melittin insertion into bilayers. The modulation of melittin association by chain packing could also rationalize the enhanced melittin binding observed for fluid-phase membranes, even those with a large proportion of demethylated ammonium group. It was previously shown that melittin binding to DOPC (34) or 1,2-dielaidoyl-sn-glycero-3-phosphocholine (DEPC) (29) membranes in the fluid phase was comparable to that observed for their equivalent PE bilayers. The binding was hindered only when DEPE membranes were cooled to the gel phase, similar to what we observed for POPE in this work. The absence of an effect of substituting PC headgroups by PE on fluid-phase binding suggests that the tightening effect of PE manifests mostly for membranes with lipid chains that are already motionally restricted. From an energetic point of view, the relatively low cohesion of fluid-phase membranes and the adaptability of the acyl chains to fill potential voids created by the peptide favor the insertion of melittin into the hydrophobic region of bilayers. However, DPPE bilayers appear to display a very low affinity for the peptide even in the fluid phase. These bilayers may be associated with a chain packing that is too ordered even in the fluid phase to accommodate the peptide inside their hydrophobic core. The chain-ordering effect of PE is reminiscent of that of Chol (57), and it was shown that the presence of Chol in PC bilayers reduced melittin affinity for membranes (58). Such a decrease could also be associated with the high energy cost related to insertion of the peptide into the ordered hydrophobic core of these membranes.

It is worth mentioning that specific interactions, such as a cation-π interaction between the ammonium group of PC and the indole moiety of tryptophan, may also contribute to the affinity of melittin for PC-rich membranes. Molecular-dynamics simulations predicted that melittin binds to DMPC membranes with its 19-Trp near the choline moiety of a neighboring phospholipid (51). However, at this point, the impact of methylation of the interfacial ammonium groups on the strength of putative cation-π interactions is not clearly defined.

Effect of demethylated headgroups on lipid extraction

A general two-step mechanism for lipid extraction by melittin has been proposed (31, 59). First, melittin inserts at the membrane interface level mainly through hydrophobic interactions. Second, the peptide relocates deeper in the hydrophobic core, disrupting the membrane and extracting fragments of the bilayer. At least three contributions have been proposed as the driving force of this rearrangement. First, it was proposed that this relocation could occur during the fluid-to-gel-phase transition, as the gel phase has a reduced capacity for accommodating the mechanical constraints exerted by inserted melittin (31). This phenomenon could explain the bicelle formation observed upon cooling after an incubation in the fluid phase (24, 42, 60, 61, 62). Second, the relocation of bound melittin from the interface toward the hydrophobic core of the bilayer could be caused by an electrostatic repulsion between bound melittins once the cationic peptide reaches a critical density at the interface. Such a relocation could be related to the change in orientation that was recently observed in molecular-dynamics simulations by Sun et al. (63). These authors proposed that the interpeptide electrostatic repulsion contributes to the chemical potential of the peptide and, at a critical peptide concentration, helps to drive the peptide from the flat part of bilayers to the membrane interior, leading to the formation of large pores that include lipids and melittin molecules. An analogous phenomenon was proposed to rationalize a significant increase in the lipid extraction susceptibility of membrane by melittin when positively charged 1,2-dipalmitoyl-3-trimethylammonium-propane (DPTAP) was introduced into DPPC membranes (31). The electrostatic repulsion between DPTAP and melittin would make the interfacial location of melittin unfavorable from a thermodynamics point of view and thus promote deeper insertion of melittin into the bilayers (31). Third, it was proposed that the melittin insertion stretches the interface area of the membrane and induces a local thinning of the bilayer (59). Above a critical peptide concentration, the internal membrane tension caused by the thinning has to be loosened by the relocation of the peptide, leading to the formation of pores and ultimately of bicelles.

Our results demonstrate that phospholipid headgroup methylation is a determining factor in lipid extraction. Phospholipids with demethylated and PE headgroups form membranes that are more resistant to melittin-induced lipid extraction. The extent of lipid extraction appeared to be dependent on the average number of methyls per ammonium group. For example, the lipid extraction obtained with DPMe₂PE (two methyls/headgroup) was comparable to that observed for DPPC/DPPE 50/50 (1.5 methyls/headgroup) and DPPC/DPPE 75/25 (2.25 methyls/headgroup) bilayers. This observation suggests that melittin-induced lipid extraction was controlled by general cohesion properties of the membranes conferred by their level of headgroup methylation, rather than by a specific molecular recognition of the PC headgroup. We have shown that the demethylation of DPPC (chemically or by the addition of DPPE) reduces melittin association with membranes in the fluid phase, likely by increasing membrane cohesion. The lower amount of active (bound) melittin naturally contributes to the limited lipid extraction. As an ultimate case, melittin does not bind to DPPE bilayers, and, indeed, no lipid extraction was observed. This finding is reminiscent of previous reports that the presence of Chol in DPPC membranes inhibits melittin-induced lipid extraction (27, 28). It was proposed that tight lipid packing due to high Chol concentrations inhibits lysis by preventing melittin penetration into membranes during incubation. It should be pointed out, however, that peptide binding to membranes and peptide-induced lipid extraction involve different intermolecular interactions, and the behavior of one cannot simply be extrapolated to the other. For example, the behavior of DPPC/DPPE 75/25 membranes appeared to be peculiar since even though less peptide was bound, the observed proportion of extracted lipids was consistently larger for these membranes than for DPPC membranes after incubation at low M/L (<3 × 10⁻³) or L/M > 333 (Fig. 6). We propose that DPPC/DPPE 75/25 bilayers have an interfacial cohesion that allows considerable melittin binding in the fluid phase, but gel-phase formation triggers a relocation of melittin and the formation of small peptide-lipid particles. Figs. 1 and 2 show that the association of melittin with DPPC/DPPE 75/25 bilayers was similar to that with DPPC ones in the fluid phase. However, the cooling resulted in a significant level of melittin dissociation from DPPC/DPPE 75/25 bilayers, likely because of a significantly reduced capability of these gel-phase bilayers to accommodate the peptide at their interface level. Therefore, the formation of a gel phase in DPPC/DPPE 75/25 bilayers could lead to a destabilization of melittin inserted into the membrane interface; the peptide would then relocate parallel to the bilayer normal and could cause increased lipid extraction. DPPC gel-phase bilayers could, in these conditions, accommodate such a limited amount of bound melittin. This phenomenon is reminiscent of the lipid extraction increase observed for DPPC membranes with 10% DPTAP in a previous study (31). This cationic lipid induced an overall reduction in the affinity of positively charged melittin for bilayers. However, its presence increased the extent of lipid extraction, and only small lipid-melittin particles were observed at an L/M ratio as high as 100. It was proposed that melittin adsorption to membranes is driven by hydrophobic interactions, but electrostatic repulsion between the positively charged bilayer interface and the cationic peptide disfavors the interfacial position and leads to melittin redistribution in the bilayers, causing lipid extraction. The cohesion of bilayers that contain a limited amount of PE could act similarly to these electrostatic interactions.

Extraction selectivity toward PC

The results clearly show that melittin selectively extracted PC from PC/PE mixed membranes: the proportion of extracted PC was two to six times greater than the extracted PE depending on the composition of the original bilayers. These findings were obtained for DPPC/DPPE as well as POPC/POPE bilayers. We propose that this lipid extraction specificity of PC over PE is a consequence of a putative depletion of PE in the vicinity of bound melittin. As described above, PEs form membranes with a tighter phospholipid packing than PCs, notably because of their capacity to form intermolecular hydrogen bonds. The presence of melittin at the membrane interface level is expected to lead to a depletion of PE in the vicinity of the peptide to maintain PE-PE contacts. Melittin molecules would subsequently extract a bilayer patch that would be PE depleted. The suggested PE depletion of the surroundings of the membrane-bound melittin is consistent with previous conclusions based on pore formation by the peptide. It was proposed that PE was excluded from melittin-rich domains that led to toroidal pores, as the curvature properties of these lipid species were unfavorable for the formation of such curved structures (34). Such a mechanism could also provide a rationale for the reported PC specificity in melittin-induced lipid extraction from DPPC/Chol membranes (27). Chol has an ordering effect on lipid acyl chains and would be excluded from the surroundings of adsorbed melittin at the bilayer interface. When the local concentration of interfacial melittin reaches a critical value, the relocation of melittin deeper in the bilayer causes the extraction of a bilayer patch that is, in these conditions, depleted in Chol relative to the sterol content of the original membrane. The proposed model for rationalizing the lipid specificity of the extraction of bilayer patches by melittin is reminiscent of a mechanism previously suggested for the action of Triton X-100 on Chol-containing membranes (64, 65). When Triton X-100 was added to vesicles made of POPC, sphingomyelin (SM), and Chol (a lipid system that formed homogeneous bilayers), it induced the formation of liquid-ordered and liquid-disordered domains due to a segregation of SM and Chol. It was found that when a sufficient amount of detergent was added, it would selectively solubilize the liquid-disordered domains, providing a POPC-enriched efflux. Therefore, it is suggested that the modulation of lipid-mixing properties in bilayers could be a simple way to control the specificity of lipid efflux from membranes.

In this work, we have demonstrated that an amphipathic helical peptide can cause lipid efflux from membranes in a lipid-selective manner, based on the methylation level of the headgroup in this case. This phenomenon is likely not limited to melittin and may play a pivotal role in other biological events. For example, previous studies established that ApoA1 specifically extracts PC over SM and Chol from mouse macrophages (66, 67, 68) and from fibroblasts (69). In these cases, the overall lipid efflux mechanism is complex, involving notably the ATP-binding cassette transporter (ABCA1) present in macrophages. To account for the selective PC extraction by Apo-A1, it was pointed out that ABCA1 is primarily found in fluid, SM- and Chol-poor regions of the membranes. The lipids that were transferred from ABCA1 to Apo-A1 lipoparticles originated from the PC-rich vicinity of ABCA1. In that system, the specificity of the lipid efflux was also proposed to be associated with a phase separation in membranes. The specificity regarding PC versus phosphatidylethanolamine (PE) is not clearly established for this system. A PC enrichment was observed in lipoparticles formed with ApoA1 incubated with fibroblast membranes (69), whereas ApoA1 incubated with POPC/POPE MLVs preferentially formed PE-enriched, saddle-shaped particles (70), a phenomenon that is associated with the polymorphic propensities of POPE. The rapid development of lipidomics is expected to reveal further specific lipid efflux caused by peptides and proteins, and it will be necessary to gain an understanding of such processes to determine the mechanism of related biological events.

Interestingly, it was recently found that styrene-maleic acid copolymers solubilized membranes into nanodiscs, a process that is reminiscent of bicelle formation by melittin. The organization of these self-assemblies is actually similar, with a small discoidal lipid bilayer whose edges are coated by the amphiphilic molecules (24, 25, 26, 71). The proposed mechanism for nanodisc formation by the copolymer is also similar to the one associated with the bicellization process induced by melittin: the copolymer first binds the interface, inserts into the membrane, and then extracts a piece of the lipid bilayer to form nanodiscs (71). Like melittin, the copolymer displayed a reduced ability to form nanodiscs with PE-containing membranes compared with DOPC membranes. However, no selectivity in lipid extraction was observed, since the extents of PE and PC extraction were the same. This discrepancy with the bicellization process caused by melittin indicates that some interactions that impact the lipid extraction induced by molecular species such as peptides, proteins, detergents, and polymers are not yet understood. This work shows that molecular details (in this case, the methylation level of the lipid headgroup) are sufficient to tune the susceptibility of membranes to lipid extraction, resulting in a selective extraction.

Author Contributions

M.L. and A.T. designed the project, analyzed data, and wrote the manuscript. A.T. performed experiments.

Editor: Amitabha Chattopadhyay.