The Lipid Dependence of Melittin Action Investigated by Dual-Color Fluorescence Burst Analysis

Abstract

Dual-color fluorescence-burst analysis was used to study melittin-induced leakage of macromolecules from liposomes of various lipid compositions. To perform dual-color fluorescence-burst analysis, fluorescently labeled size-marker molecules were encapsulated into liposomes, labeled with a second lipid-attached fluorophore. By correlating the fluorescence bursts, resulting from the liposomes diffusing through the detection volume of a dual-color confocal microscope, the distribution of size-marker molecules over the liposomes was determined. It was found that melittin causes leakage via two different mechanisms: 1), For liposomes composed of neutral bilayer-forming lipids, low melittin concentrations induced pore formation with the pore size depending on the melittin concentration. 2), For liposomes containing anionic and/or nonbilayer forming lipids, melittin induced fusion or aggregation of liposomes accompanied by a-specific leakage. Experiments with liposomes prepared from Escherichia coli lipid extracts and intact cells of Lactococcus lactis indicate that both mechanisms are physiologically relevant.

INTRODUCTION

Antimicrobial peptides are typically 12–50 amino acids long and play an important role in the innate immune response of the host organism. Over 800 antimicrobial peptides have been identified in species as diverse as bacteria, fungi, plants, and mammals (1) and they are a possible source of antibiotics for the pharmaceutical industry (2,3). The antimicrobial peptide melittin is a membrane disrupting agent, frequently used as a model for hemolytic activity (4) and is one of the best studied antimicrobial peptides to date (see Raghuraman and Chattopadhyay (5) for recent review). The strongly cationic (6+) melittin is the major component of the European honeybee (Apis mellifera) venom (6) and consists of 26 residues. In aqueous solution, melittin exists as either a monomer, with no defined secondary structure (7,8), or as a tetramer of α-helical peptides, depending on the concentration, ionic strength, and pH of the solution (8–10). It retains the amphiphilic α-helical structure in the membrane (11–13). Melittin has strong hemolytic and antimicrobial activity. It kills the cell by autoinsertion in the membrane, thereby forming large a-specific pores in both eukaryotic and prokaryotic membranes, leading to the leakage of cell content (14). Other cytotoxic mechanisms of action have also been proposed for melittin, including phospholipase D activation that leads to hydrolysis of phospholipids and probably activation of unknown transduction cascades (1,5,14–18).

It has been proposed that melittin induces pore formation via the toroidal mechanism (19,20), whereby the lipids, in particular the headgroups, together with the peptides line the pore. However, the toroidal mechanism is not undisputed and evidence is also available in favor of a barrel-stave arrangement of the melittin peptides in the pore (13,21–24). In the barrel-stave arrangement, the amphiphilic α-helical peptides insert in the membrane bilayer in such a way that the hydrophilic regions, but not the lipid headgroups, form the interior of the pore. There are also reports that claim that melittin disrupts membranes via the carpet-like mechanism (4,22). In the carpet-like mechanism, the peptides disrupt the membrane by forming an extensive layer of membrane-associated peptides and dissolve the membrane in a detergent-like manner with the helices orienting parallel to the surface. According to the widely accepted toroidal pore model, the pores consist of approximately four to eight melittin molecules (25,26). Regardless of the lipid composition, at low peptide/lipid ratios, the helical segments bind parallel to the membrane plane (27–31), at the depth of the glycerol groups (32). Each peptide would occupy an area of ∼6 nm² of membrane surface (33). Depending on the membrane composition, an increase in melittin concentration is thought to result in reorientation of melittin and insertion of the peptide perpendicular to the membrane plane (20), which induces the formation of pores (4,13,34).

The size of the melittin pore has been estimated to be between 3.5 and 4.5 nm in transmission electron microscopy (35) and 4.4 nm in neutron diffraction studies (20). This is in reasonable agreement with estimates from leakage experiments of macromolecules (25,34–41), where pore sizes ranging from 1.5 to 4.8 nm have been estimated. It should be noted, however, that the difference between the smallest (1.5 nm) and the largest pore-size estimate (4.8 nm) corresponds to molecules diffusing through the pore of <10 kDa up to >100 kDa and this apparent discrepancy might be of crucial importance for the mechanism of cell killing by melittin. It seems likely that these differences in pore-size estimates arise from differences in salt concentration, osmotic gradient, and lipid composition (1,5,42,43). For a better understanding of the action mechanism and the cytotoxic activity of melittin, a better knowledge of the pore formation and pore size is essential. Whether and how the pore size increases with increasing melittin concentration is still a matter of debate (20). In previous studies, phospholipid vesicles and erythrocyte membranes were used to monitor the leakage of macromolecules as a function of the melittin concentration, but the mechanism of pore formation and size could not be specified unambiguously (34–36,44).

We used dual-color fluorescence-burst analysis ((DCFBA); (45)) to probe the formation and size of the melittin pore. DCFBA allows one to determine the population distribution of size-marker molecules over the liposomes, thereby providing detailed information of the leakage process that is not readily obtained by conventional methods. Moreover, only small quantities of lipids are needed (∼μg) and the data acquisition is relatively fast (∼10 min), which allows for medium throughput screening of lipid compositions and melittin concentrations.

METHODS

Melittin and lipids

Melittin was obtained from GenScript (purity >95%; Piscataway, NJ). Escherichia coli total lipid extract, cholesterol, and the synthetic lipids 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE), and 1,2-dioleoyl-sn-glycero-3-phosphatidylglycerol (DOPG) were from Avanti Polar Lipids (Alabaster, AL). Stock solutions of the lipids of 20 mg ml⁻¹ in chloroform were prepared. The membrane dyes DiO (3,3′-dioctadecyloxacarbocyanine percholate; excitation wavelength λ_Ex = 484 nm, emission wavelength λ_Em = 501 nm) and DiD (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine percholate; λ_Ex = 644 nm, λ_Em = 665 nm, Invitrogen, Carlsbad, CA) were added to the lipid stock solutions to ensure coincorporation in the vesicle membrane. For DOPC, a 1:4000 DiO (or DiD)/lipid ratio was used; for all other lipid mixtures, a 1:40,000 ratio was used. To obtain similar fluorescence-burst intensities, the probe/lipid ratio had to be 10-fold higher for DOPC. This is probably due to different interactions of DiO with zwitterionic DOPC when compared to the anionic DOPG. Experiments with liposomes composed of pure DOPG or a 1:1 mixture of DOPC/DOPG indicated that a 10-fold higher DiO concentration did not influence the leakage and fusion events (not shown).

Labeling of size-marker molecules and preparation of liposomes

Dextran molecules conjugated to Alexa fluor 647 (for the 10 kDa dextran, λ_Ex = 650 nm, λ_Em = 668 nm) and Oregon green 488 (for the 70 kDa dextran, λ_Ex = 496 nm, λ_Em = 521 nm) were from Invitrogen and dissolved in 50 mM potassium phosphate, pH 7.0, to a final concentration of 100 μM. The effective Stokes radii of the 10 and 70 kDa dextrans are 2.4 and 5.0 nm, respectively (36,46), although in solution they form ellipsoids with shortest radii of ∼2.0 nm (47). Glutathione, bovine pancreatic trypsin inhibitor (BPTI), and bovine α-lactalbumin were from Sigma-Aldrich (St. Louis, MO). Solutions of 100 μM BPTI and α-lactalbumin in 30 mM K-Hepes, pH 8.6, were labeled with 44 μM Alexa fluor 633 C5-maleimide (Invitrogen, λ_Ex = 622 nm, λ_Em = 640 nm, Mw = 1089 (45)) by incubating for 30 min at room temperature. The labeled macromolecules were separated from the unbound fluorophore by size exclusion chromatography (Nap10 columns, Amersham Biosciences, Piscataway, NJ) with subsequent buffer exchange to 50 mM potassium phosphate, pH 7.0. Glutathione was labeled with the same fluorophore in 50 mM potassium, pH 7.0. The purity of the proteins and efficiency of labeling was checked by SDS-PAGE and in-gel fluorescence.

For the liposome preparation, 250 μl (5 mg) of the lipid chloroform stock was dried by rotary evaporation at 50°C for 2 h. The lipid film was resuspended in 1 ml of 50 mM potassium phosphate, pH 7.0, with the size-marker molecules present at concentrations between 1 and 20 μM. After 3 times fast freezing (liquid nitrogen) and slow thawing (room temperature) and subsequent extrusion through 200 nm polycarbonate membranes (Avestin, Ottawa, Canada), the nonencapsulated size-marker molecules were separated from the liposomes by centrifugation (20 min, 270,000 × g, 4°C). For the DCFBA measurements, the liposomes were resuspended in 50 mM potassium phosphate, pH 7.0, to a final concentration of 80 μg of lipid ml⁻¹.

DCFBA measurements

DCFBA measurements were performed as described previously (45). Briefly, the fluorescently labeled size-marker molecules were encapsulated in liposomes that were labeled with a second fluorophore (the lipid analog DiO or DiD). The fluorescent bursts, resulting from the diffusion of the liposomes through the detection volume of a dual-color confocal microscope (45), were recorded for at least 10 min (Fig. 1). During this period, >1000 liposomes passed through the focal volume (Fig. 2). For all the fluorescent bursts above a certain offset, the integrated intensities of both the fluorescent lipid analog and the size-marker molecules were calculated. The size-marker concentration (C) of the i^th liposome (burst) could then be obtained by division:

(1)

where the fluorescence of the lipid marker (I_L) is above a certain offset between t₂ and t₁ and I_SM is the fluorescence of the size marker. In Eq. 1, the signals are normalized because the lipid analog is associated with the surface of the liposomes and the size-marker is present in the vesicles. As a function of liposome radius, I_SM and I_L thus scale to the power of 3 and 2, respectively (45). From C_i, the average concentration C_av over all the liposomes could be calculated:

(2)

where N corresponds to the number of liposomes (45).

FIGURE 1.

Principle of dual-color fluorescence-burst analysis. Liposomes labeled with the fluorescent lipid analog DiO and with encapsulated Alexa fluor 633 (AF633)-labeled size-marker molecules are depicted schematically. The fluorescent bursts, resulting from the diffusion of liposomes through the detection volume of a dual-color confocal microscope, are recorded for >10 min. Because both DiO and Alexa fluor 633 are liposome associated, the intensity bursts in both channels coincide. Addition of melittin results in pore formation and pore-size dependent leakage of the size-marker molecules from the liposomes and thereby, in turn, cancels out the coincidence of the fluorescence bursts.

FIGURE 2.

Fluorescence traces from DCFBA experiments. The upper panels show the intensity traces for the fluorescent lipid analog DiO and the lower panels the corresponding traces of the encapsulated Alexa fluor 633-labeled glutathione. Traces are shown for liposomes consisting of: DOPC without (A) and with melittin ((B) the melittin/lipid ratio was 0.03, which corresponds to a concentration of 4 μM), and DOPG without (C) and with melittin ((D) the melittin/lipid ratio was 0.09, which corresponds to a concentration of 12 μM). Melittin eliminates the bursts from Alexa fluor 633-labeled glutathione, because glutathione is no longer associated with the liposomes and the signal from individual molecules is low. Only the first 25 s of >10 min are shown.

RESULTS

Pore formation

The effect of the lipid composition on the melittin pore size was studied using DCFBA, which enables rapid measurements of leakage with single liposome resolution (45). The liposomes were labeled with the fluorescent lipid analog DiO or DiD and encapsulated with size-marker molecules that were labeled with a second fluorophore. Because both the fluorescent lipid analog and the size-marker molecules are liposome associated, both fluorescence signals coincide (Fig. 1, top panel). Upon pore formation, the size-marker molecules with dimensions smaller than the diameter of the melittin pores leak out (Fig. 1, bottom panel). This results in the loss of coincidence of the signals. In Fig. 2 A, part of a typical intensity trace for liposomes consisting of 100% DOPC with encapsulated fluorescently labeled glutathione is shown. After addition of melittin, glutathione leaked out of the liposomes and the fluorescence bursts no longer coincided (Fig. 2 B). The same phenomenon was observed for liposomes composed of the anionic lipid DOPG (Fig. 2 C–D). Each individual burst from the fluorescent lipid analogs corresponds to a liposome diffusing through the detection volume, and, for such a liposome, the concentration of size-marker was calculated (Eq. 1). Because traces were recorded for longer than 10 min, the concentration distributions of large populations (typically >1000 liposomes) were obtained. Steady-state levels of efflux were reached, since the concentration distributions were stable for >24 h (not shown).

Fig. 3 A presents the concentration distribution at different melittin concentrations of DiO-labeled liposomes composed of DOPC with encapsulated Alexa fluor 633-labeled glutathione (Eq. 1). Upon addition of melittin, pores were formed and the glutathione leaked out, resulting in lower internal concentrations. Liposomes composed of DOPG showed a similar behavior, albeit at higher melittin concentrations (Fig. 3 B). Surprisingly, the number of liposomes decreased with increasing melittin concentration (the surface of the histogram corresponds to the concentration of liposomes). For liposomes composed of a mixture of DOPC/DOPE, both the size-marker concentration and the total number of liposomes dropped already at low melittin/lipid ratios (Fig. 3 C). This decrease of the liposome concentration was due to liposome fusion or aggregation, since the average DiO burst intensities increased in parallel (Fig. 4 A). Moreover, bright fluorescent micrometer-size particles were observed in microscopic images, which were not visible before the addition of melittin (Fig. 4 B). The detergent Triton X-100 (used at 0.05% v/v) fully dissolved the liposomes and completely abolished the bursts of both the fluorescent lipid analogs and the size-marker molecules (not shown).

FIGURE 3.

Pore formation of melittin. The distribution of Alexa fluor 633-labeled glutathione ((C) Eq. 1) over the liposomes is shown for various melittin/lipid ratios (ratios are indicated in the figure). The distributions are shown for liposomes consisting of: (A) DOPC, (B) DOPG, and (C) a 3:7 ratio of DOPC/DOPE. Upon the addition of melittin, the concentration of glutathione decreased due to leakage from the liposomes. Importantly, for DOPG and DOPC/DOPE the total number of liposomes (area of the histogram) also decreased upon the addition of melittin due to membrane fusion or aggregation. The bin width used was 0.16.

FIGURE 4.

Melittin-induced fusion of liposomes consisting of 100% DOPG. (A) Fluorescence intensity traces of the fluorescent lipid analog DiO before (upper graph) and after the addition of a melittin (lower graph). Upon the addition of melittin (melittin/lipid ratio of 0.09; 12 μM), the number of fluorescence bursts (related to the liposome concentration) decreased and the average burst intensity increased. (B) Fluorescence microscopy images of the samples from panel A. After addition of melittin, micrometer-size particles became visible.

Lipid composition dependence

To quantify the melittin activity for different lipid compositions, the total number of liposomes was plotted as a function of the melittin concentration. Typical curves are shown in Fig. 5 (solid square) for liposomes consisting of 100% DOPC (A), 100% DOPG (B), and a 3:7 ratio of DOPC/DOPE (C). Also, the average concentration of encapsulated size marker was calculated (Eq. 2), and for Alexa fluor 633-labeled glutathione, these are also shown in Fig. 5 (solid circle). For DOPC, glutathione leakage was observed already at very low melittin concentrations, whereas membrane fusion was not observed below melittin/lipid ratios of 0.4 (Fig. 5 A; only data up to 0.08 are shown). For DOPG, glutathione leakage was observed at concentrations about 10 times higher than for DOPC (Fig. 5 B). Interestingly, liposome fusion or aggregation was observed in the same concentration range as the leakage of contents. Similar behavior was observed for liposomes consisting of a 3:7 ratio of DOPC/DOPE (Fig. 5 C), albeit at significantly lower melittin concentrations than for DOPG (notice the differences in the scales of the x axes of Fig. 5, B–C). To obtain C_L,1/2, defined as the concentration of melittin where half of the liposomes released their content, the size-marker concentration curve (Fig. 5, solid circle, solid line) was fitted with a normal cumulative distribution function:

(3)

where C_av is the average size-marker concentration of the liposomes normalized to the condition where no melittin was present, C_M is the melittin concentration, σ² is the variance, and erf indicates the error function. Similarly, the melittin C_F,1/2 concentrations at which half of the liposomes were fused or aggregated (Fig. 5, solid square, dotted line) were calculated using Eq. 3, with C_av corresponding to the normalized number of liposomes.

FIGURE 5.

Melittin titration curves. Normalized, average glutathione concentrations (C_av, Eq. 2, •) and total number of liposomes (▪) plotted as a function of the melittin/lipid ratios for liposomes composed of: (A) DOPC, (B) DOPG, and (C) a 3:7 ratio of DOPC/DOPE. The data were normalized to the conditions where no melittin was present. For the total number of liposomes, the data sets for the four different size markers were averaged and the error bars show the standard deviation. The solid and dotted lines are fits with a cumulative distribution function (Eq. 3), used to obtain the melittin/lipid ratio where leakage or fusion was 50% (C_L,1/2- and C_F,1/2-values, respectively). For liposome fusion or aggregation, the C_F,1/2-values with spread σ² were as follows: (A) C_F,1/2 = 0.3, σ² = 0.9 (40 μM), (B) C_F,1/2 = 0.07, σ² = 0.33 (9.3 μM), and (C) C_F,1/2 = 0.005, σ² = 0.001 (0.7 μM). For glutathione leakage, the C_L,1/2-values were: (A) C_L,1/2 = 0.002, σ² = 0.001 (0.3 μM), (B) C_L,1/2 = 0.05, σ² = 0.15 (6.7 μM), and (C) C_L,1/2 = 0.005, σ² = 0.002 (0.7 μM).

The C_F,1/2-values for fusion or aggregation were determined for a number of lipid compositions. DOPC fused at melittin/lipid ratios of 0.44 ± 0.04. Addition of 10% cholesterol reduced the melittin/lipid ratio needed for fusion or aggregation to 0.25 ± 0.07 (Fig. 6 A). Liposomes consisting of 1:1 and 3:7 ratios of DOPC/DOPE fused or aggregated at melittin/lipid ratios of 0.13 ± 0.03 and 0.004 ± 0.003, respectively. DOPE (100%) was not tested, since it did not form stable bilayers under the used experimental conditions (48). Similar to DOPE, DOPG decreased the amount of melittin needed to induce liposome fusion or aggregation, and, for 100% DOPG, this was induced at a melittin/lipid ratio of 0.06 ± 0.02. Analogous to liposome fusion or aggregation, the C_L,1/2-values were determined for leakage of different size markers (Fig. 6 B). Fluorescently labeled glutathione (1.4 kDa), BPTI (7.6 kDa), 10 kDa dextran (10 kDa), and α-lactalbumin (15.3 kDa) were used as size markers. In the case of 100% DOPC and DOPC with 10% cholesterol, the smaller size-marker molecules leaked out at lower melittin concentrations than the bigger ones, indicating a pore-size dependence on the melittin concentration. For liposomes containing synthetic DOPE or DOPG this size dependence was abolished (Fig. 6 B), suggesting that leakage is a consequence of the fusion or aggregation of the liposomes rather than formation of specific pores. The 70 kDa dextran was also tested for liposomes consisting of 100% DOPG or a 1:1 mixture of DOPC/DOPE, and this compound leaked out at similar melittin concentrations as the smaller size markers (not shown). Thus, in membranes with a fraction of DOPE or DOPG, melittin triggers the formation of holes with sizes of at least 2.3–5 nm (size of the largest size markers (45,46)). Interestingly, for liposomes consisting of E. coli total lipid extract, there was a melittin concentration dependence for the leakage of the three smallest size markers (Fig. 6 B), and hence a pore-size dependence. E. coli membranes consist of anionic lipids (phosphotidylglycerol (PG) plus cardiolipin) and phosphatidyletanolamine (PE) in a ratio of ∼3:7 (49). The C_F,1/2-value for fusion or aggregation, but not the C_L,1/2-value for leakage, was similar for membranes consisting of a 3:7 mixture of DOPG/DOPE (0.03 ± 0.01, Fig. 6 A).

FIGURE 6.

Melittin action on liposomes of various lipid compositions. (A) C_F,1/2-values for liposome fusion or aggregation, reflecting the melittin/lipid ratio where half of the liposomes were fused or aggregated. The bars represent the standard deviation for >10 measurements. (B) C_L,1/2-values for leakage of the fluorescent size-marker molecules. The size markers are arranged on increasing size: GSH, Alexa fluor 633-labeled glutathione (molecular mass, 1.4 kDa (45)); BPTI, Alexa fluor 633-labeled bovine pancreas tryptic inhibitor (7.6 kDa (45)); 10 kDa dextran, Alexa fluor 647-labeled dextran molecule (10 kDa); α-lactalbumin, Alexa fluor 633-labeled α-lactalbumin (15.3 kDa (45)).

In vivo leakage experiments

DCFBA is based on the fluorescent labeling of liposomes and encapsulated size-marker molecules labeled with a second fluorophore. Here, we demonstrate that this approach can also be used to probe pore formation in bacteria instead of liposomes, using GFP as a leakage marker. Lactococcus lactis (strain MG1363) cells, expressing the OpuC-GFP fusion protein from a plasmid, were labeled with the red membrane probe DiD. OpuC is the 24 kDa substrate-binding domain of OpuA (an osmoregulated ABC transporter from L. lactis), expressed as a soluble protein (50). OpuC-GFP can be considered as two beads each with a Stokes radius of 2.8 nm (51). The inset of Fig. 7 shows confocal images of L. lactis cells expressing OpuC-GFP (upper image) and labeled with DiD (lower image). The main figure presents the GFP concentration before and after 5 min treatment with 4 μM melittin. As can be seen from the figure, melittin caused leakage of OpuC-GFP and thus formation of pores of at least 2.8 nm.

FIGURE 7.

In vivo leakage experiment. DCFBA experiments were performed on L. lactis cells expressing a OpuC-GFP fusion protein of 51 kDa (inset, bottom panel). The cell membranes were stained with the red fluorescent lipid analog DiD (inset, top panel). The distribution of the GFP-fusion protein over the cells is presented in the main figure (▪) After addition of 4 μM melittin, the GFP-fusion protein leaked from the cells and this resulted in a decrease of the internal GFP concentration (•).

DISCUSSION

In the case of membranes prepared from 100% DOPC, Alexa fluor 633-labeled glutathione leaked out at melittin/lipid ratios of 0.002 ± 0.001, which is consistent with literature values for liposomes with PC headgroups and small size-marker molecules (20,34–36). Solid-state NMR studies have shown that melittin has a transmembrane orientation at these melittin/lipid ratios (22–24). For larger size-marker molecules, leakage occurred at higher melittin concentrations (Fig. 6 B), demonstrating that the pore size is related to the melittin concentration. It has been reported that at melittin/lipid ratios of <0.01, the fluorophore calcein leaked out, but not labeled dextran (Stokes radius of 1.3 and 2.4 nm, respectively), whereas at higher melittin concentrations both leaked out (36). Other studies also provide evidence for a certain relationship between the pore size and melittin concentration, since large size-marker molecules leaked to a lesser extent or at a higher melittin concentration than the smaller ones (34,35). At high melittin/lipid ratios of 0.44 ± 0.04, melittin fused or aggregated liposomes (Fig. 6 A). At this melittin/lipid ratio, the phospholipids structure is still that of a bilayer with minimal disruption (52,53). The fusogenic effect of melittin has been reported in the literature (35,54–56) but is now quantitatively differentiated from the formation of discrete pores. The presence of 10% cholesterol increased the melittin/lipid ratios needed for efflux, relative to 100% DOPC liposomes (Fig. 6 B), and the activity of melittin was reduced. It is well known that cholesterol reduces the pore-forming activity of melittin (5,9,26,43,57) and this is probably due to reduced binding of the peptide to the bilayer (58–60). The reduced melittin activity may also be accounted for by tight lipid packing and increased deformation energy induced by cholesterol (5).

Melittin was much less active in liposomes containing a fraction of DOPG than in liposomes of 100% DOPC liposomes (Fig. 6 B), whereas the absorption to the liposome surface of the cationic peptide is approximately sixfold higher (33). For 100% DOPG liposomes, melittin/lipid ratios of 0.045 ± 0.001 were needed to induce leakage of glutathione, and this is ∼25 times higher than that for 100% DOPC liposomes. For liposomes consisting of a 1:1 ratio of DOPC/DOPG, an intermediate melittin/lipid ratio of 0.028 ± 0.001 was needed to induce leakage (Fig. 6 B). It is well known that melittin is less active with negatively charged lipids like DOPG than with neutral lipids (26,35–42,61). This phenomenon has been explained by the difficulty in membranes with a negative surface charge to insert the melittin α-helix, which requires a reorientation of the molecule from parallel to perpendicular to the membrane plane. In membranes with negatively charged headgroups, the cationic melittin is more tightly bound in comparison to membranes with zwitterionic PC headgroups (4). This proposal is supported by the finding that PC membranes have a lower capacity for surface absorption of melittin in comparison to negatively charged bilayers (33).

Importantly, for DOPG-containing liposomes, there was no difference in melittin activity for the leakage of different size markers (Fig. 6 B). Thus, contrary to 100% DOPC liposomes, there was no relationship between the pore size and the melittin concentration and the holes formed must have been at least 3.2–5 nm in diameter (size of α-lactalbumin and 70 kDa dextran (45,46)). This is in agreement with the finding that PG-containing membranes have no preference for 4 and 70 kDa dextrans, whereas more of the smaller compound leaked out from zwitterionic PC membranes (4). Another important difference was that melittin induced fusion or aggregation of DOPG containing liposomes at similar concentrations as leakage, whereas for 100% DOPC liposomes, this concentration was 100–1000 times higher (Fig. 6). These findings are in accordance with turbidity measurements, where the size of small unilamellar vesicles increased for negative lipids, but not for neutral lipids (35,55). It has been suggested that melittin does not insert into PG-containing bilayers, but only accumulates at the surface and that this asymmetric accumulation eventually leads to nonselective membrane damage and efflux via nonspecific holes (4,58). This is supported by the finding that aggregation of surface-bound α-helical melittin leads to a reduced bilayer thickness (62–65) and consequently local structural instabilities and fluctuations (4). According to our findings, these nonselective pores coincide with fusion or aggregation (Fig. 5 B). Thus, in negatively charged membranes, melittin aggregates either in the membrane or on the membrane surface, which leads to fusion or aggregation of the lipid bilayers, accompanied by nonspecific contents leakage from the liposomes. These morphological changes of membranes induced by melittin occur only when the lipids are in the crystalline liquid phase and the hydrocarbon chains have a disordered conformation (32). It is well known that if the lipids are in the gel-phase, melittin induces micellization of the vesicles and, at high concentrations, a complete loss of vesicular structure via the carpet-like mechanism (31,32,37,57,66–69).

The presence of 50% DOPE had similar effects on the melittin activity as DOPG. Higher melittin concentrations were necessary to induce leakage of the smaller size-marker molecules from liposomes composed of a 1:1 mixture of DOPC/DOPE, compared to 100% DOPC. Also, all size-marker molecules leaked out at similar melittin concentrations and the pore-size dependency on the melittin concentration was lost (Fig. 6 B). DOPE is a nonbilayer forming lipid and promotes a negative membrane curvature (48). It is known that lipids of positive curvature (lysoPC) facilitate and lipids with negative curvature (PE) inhibit melittin pore formation (9,19,35). Lipids of positive curvature facilitate the bending of the membrane into the toroidal pore. Additionally, melittin binds with a 10–100-fold lower dissociation constant to membranes composed of PE/PG than to PC/PG (58,70), and this might prevent the pore formation. For liposomes consisting of a 1:1 ratio of DOPC/DOPE, fusion or aggregation occurred at melittin/lipid ratios of 0.13 ± 0.03 and this is approximately threefold lower than for 100% DOPC liposomes. This difference is probably due to the destabilizing effect of bilayers by PE (48), an effect that is opposed by melittin (53). A DOPE content of 70% resulted in liposome fusion or aggregation already at 0.004 ± 0.003 melittin/lipid ratio. The melittin concentrations needed for membrane fusion and contents leakage were much lower for liposomes consisting of a 3:7 ratio of DOPC/DOPE than for liposomes consisting of the same ratio of DOPG/DOPE (Fig. 6).

In conclusion, melittin has two different effects on membranes (Fig. 8): i), pore formation with the pore-size dependent on the melittin concentration; and ii), membrane fusion or aggregation accompanied by nonspecific leakage. The first effect was observed at low melittin concentrations for liposomes consisting of 100% DOPC or 90% DOPC with 10% cholesterol. The second effect dominated in all lipid mixtures containing the anionic lipid PG or the nonbilayer lipid PE. Interestingly, for liposomes consisting of E. coli total lipid extract, a melittin concentration dependence was observed for leakage of the three smallest size markers (Fig. 6 B). This was not observed for liposomes composed of a 3:7 ratio of DOPG/DOPE, which can be regarded as a mimic of the E. coli total lipid extract (49). This difference must be due to other factors, like acyl chain length, degree of fatty acid saturation, or the presence of tracer amounts of other lipids, which may cause a different lipid packing and hence binding of melittin. Whatever the basis for the difference is, our measurements emphasize the importance of comparing synthetic and native-like lipid mixtures in studying the pore formation of antimicrobial peptides. For another group of antimicrobial peptides (rabbit defensins), similar differences between E. coli and synthetic lipids have been reported (38,71). It was recently shown that melittin leads to aggregation of intact Candida albicans, E. coli, and Staphylococcus aureus cells (35), and probably both the pore formation and the leaky-fusion mechanism are relevant in vivo. DCFBA can be used for an in vivo leakage assay, as was shown in preliminary experiments with L. lactis cells expressing a GFP-fusion protein (Fig. 7).

FIGURE 8.

Model for the action of melittin. In membranes composed of neutral (zwitterionic), bilayer-forming lipids, low melittin concentrations induce pore formation. The pore size is dependent on the melittin concentration. A further increase in the melittin concentration will ultimately lead to membrane fusion or aggregation. In membranes containing anionic and/or nonbilayer forming lipids, melittin addition results in nonspecific membrane leakage and fusion or aggregation.

General techniques that have been employed to study antimicrobial peptides are x-ray crystallography, and optical, fluorescence, NMR, Fourier transform infrared, Raman, and circular dichroism spectroscopy (1). DCFBA is a powerful addition to these techniques, since it allows the detection of concentration distributions and is able to discriminate leakage from membrane fusion and lysis ((45); this work). It has further advantages of being relatively fast with acquisition times of ∼10 min and using very little (∼μg) material. Therefore, DCFBA enables medium-throughput screening of lipid compositions and size-marker molecules, and we intend to use it to study different antimicrobial peptides. One could also employ DCFBA to determine effects salt and osmotic gradients on melittin action and thereby obtain a better understanding of the in vivo effects (5,43).