Mechanism of Magainin 2 a Induced Permeabilization of Phospholipid Vesiclest
Earl Grant,T. Beeler,K. Taylor,K. Gable,Mark,’. Roseman
Abstract:The magainins, peptide antibiotics secreted by the frog Xenopus laevis, have previously been shown to permeabilize phospholipid vesicles. To elucidate the mechanism of permeabilization, we have conducted detailed kinetic studies of magainin 2 amide (mgn2a)hduced release of 6-carboxyfluorescein from vesicles of phosphatidylserine. The results show that dye release occurs in (at least) two stages-an initial rapid phase, with t l p = 3 s, followed by a much slower phase that approaches zero leakage rate before all the dye is released. Light-scattering studies showed that mgn2a does not cause gross changes in vesicle structure. The peptide was found to rapidly equilibrate betweenvesicles; this was demonstrated by determining a binding isotherm for the peptidelipid interaction, and by showing that addition of unloaded vesicles rapidly quenches peptide-induced leakage from loaded vesicles. Transient dye release in the presence of an equilibrating peptide can be explained in two ways: (1) the peptide exists only transiently in an active form; (2) the vesicles are only transiently leaky. Preincubation of mgn2a at assay concentrations in buffer alone or with unloaded vesicles did not inactivate the peptide. Therefore, rapid leakage is probably due to transient destabilization of the vesicle upon addition of mgn2a. The magainins, a class of antimicrobial peptides secreted fromtheskinof thefrogxenopus laevis (Zasloff, 1987; Zasloff et al., 1988), are thought to kill cells by permeabilizing the plasma membrane. A total of 23 residues in length and bearing a positive charge of +3 to +5 at physiological pH, these peptides could potentially form amphipathic helical structures that are able to disrupt lipid bilayers or form oligomeric transmembrane channels [see Williams et al. (1990) and references cited therein]. Evidence that magainins can act directly on lipid bilayers comes from studies with planar bilayers and liposomes. Duclohier et al. (1989) found that magainin 1 forms anionselective channels in planar lipid bilayers under conditions of high salt and high trans-negative potential. Several conductance states were identified, the lowest being 80 pS. The dose response curves suggested that the magainin channel consists of three to six monomers. When magainin 2 amide (mgn2a)l or its analogs were added to respiring cytochrome oxidase-containing liposomes (which have an inside-negative membrane potential), Juretic et al. (1 989) found that the peptides uncouple respiration and cause a stable (rather than transient) decrease in the steady-state membrane potential. The sigmoidal dose response curve was attributed to magainin monomers interacting cooperatively in the membrane to form an active aggregate of four to five subunits. These two studies suggest that the mechanism of magainin action is similar to that proposed for other voltage-dependent pore-formers, such as alamecithin [reviewed by Stein (1 986)]. According to the model, monomers or aggregates of the peptide initially bind to the bilayer interface (through a combination Supported by USUHS Protocol C071AH. * To whom correspondence should be addressed. Telephone: (301) t Brooke Army Medical Center. f Uniformed Services University of the Health Sciences. 295-3570. Abbreviations: mgn2a, magainin 2 amide; PS, phosphatidylserine; PE, phosphatidylethanolamine; PC, phosphatidylcholine; SUV, small unilamellar vesicle; LUV, large unilamellar vesicle; CF, 6-carboxyfluorescein; Pipes, piperazine-N,N’-bis(2-ethanesulfonic acid). of electrostatic and hydrophobic interactions) as an inactive amphipathic a-helix. The inactive form is then driven into the bilayer by the membrane potential, whereupon the peptide forms an active transmembrane pore. Changes in aggregation state can occur before, during, or after voltage-induced transmembranal insertion. There is, however, evidence to show that magainin effectively permeabilizes liposomes in the absence of a membrane potential: When magainin 1 was added to sonicated phosphatidylserine (PS) vesicles, Matsuzaki et al. (1989a) found that entrapped calcein was rapidly released, on a time scale of seconds to minutes. By measurement of the initial rate of leakage as a function of total peptide and total lipid concentration, an equilibrium binding isotherm for the peptidelipid interaction was also determined. The resulting curve, which fit well to the Langmuir equation, suggested to Matsuzaki et al. that extensive aggregation of peptide on the vesicle does not occur. From this, they concluded that magainin 1 probably does not form oligomeric channels in the bilayer, and therefore functions instead as a bilayer destabilizer. An important caveat, which they noted, is that their binding data are not sensitive enough to rule out the possibility that a small fraction of the peptide forms oligomeric channels. In our opinion, binding isotherms are exceedingly difficult to interpret because steric effects and changes in membrane surface potential lead to large activity coefficient effects that are hard to predict quantitatively (Schwarz et al., 1986). To help clarify the mechanism of magainin’s lytic action, we have studied in detail the kinetics of mgn2a-induced release of 6-carboxyfluorescein (CF) from phosphatidylserine liposomes in the absence of a membrane potential. Our results show that the fast release of dye is a transient effect. After experimentally eliminating several explanations, we conclude that fast release results from transient destabilization of the bilayer upon initial interaction with the peptide. A unifying mechanism for magainin action is proposed. EXPERIMENTAL PROCEDURES Materials. Phosphatidylserine (bovine brain) was purchased from Avanti Polar Lipids; Triton X-100, especially This article not subject to U.S. Copyright. Published 1992 by the American Chemical Society Permeabilization of Liposomes by Magainin 2 Amide purified for membrane research, was from Boehringer Mannheim. Carboxyfluorescein from Eastman (>99%) was used for most of these studies, without further purification; chromatographically pure CF (Molecular Probes) was also used in selected experiments, and gave results indistinguishable from those obtained with Eastman CF. Magainin 2 amide was a generous gift from Dr. Hao-Chia Chen. Buffers. Buffers used in SUV studies were as follows: buffer A, 10 mM Pipes, 150 mM NaCl, 1 mM EDTA, and 100 mM sucrose, pH 7.0; buffer B, 10 mM Pipes, 150 mM NaC1,’and 1 mM EDTA, pH 7.0; dye solution, 100 mM CF, 10 mM Pipes, 150 mM NaCl, and 1 mM EDTA, pH 7.0. Preparation of Small Unilamellar Vesicles ( S W s ) . Because the phase transition of the bovine brain PS is 8 “C, SUVs and LUVs had to be prepared and stored above this temperature; for convenience, all procedures were carried out at or near room temperature. Approximately 20 mg of PS in 2 mL of solvent was dried to a film in a 25-mL flask by rotary evaporation, redissolved in 5 mL of benzene, and lyophilized overnight under high vacuum. The lyophilized lipid was then suspended in 2 mL of dye solution by vortex mixing, transferred to a polycarbonate tube, and sonicated (under argon) at 25-30 OC to constant clarity in a cup horn using a Heat Systems ultrasonic sonicator (W-375). Unentrapped CF was removed from the solution by passing the sonicate through an Econo-Pac lODG column (10 mL) using buffer A as the eluting buffer. To obtain a homogeneous population of SUVs, larger particles were removed by sedimentation in an angle rotor (Ti50, Beckman ultracentrifuge), as described by Barenholz et al. (1977). Centrifugation at 45K rpm for 15 min was sufficient to produce a population of CF-loaded lipsomes that elutes as a homogeneous population of SUVs from a Sepharose 2B-CL column: A single, well-included peak was obtained, with a constant ratio of CF/phosphate across the elution profile. This result also shows that the vesicles are relatively stable, since vesicle breakdown or fusion usually produces a heterogeneous mixture of particles. Lipid phosphate was determined by the method of Bartlett (1959). Preparation of Large Unilamellar Vesicles ( L W s ) . CFloaded PS LUVs were prepared by the reverse-phase ether evaporation method (Diizgiines et al., 1983), using 100 mM CF, 10 mM Pipes, 10 mM NaC1, and 1 mM EDTA, pH 7.0 (296 mOsM), as the aqueous phase. The initially formed vesicles were extruded successively through Nucleopore filters of 0.4, 0.2, 0.1, and 0.08 pm. To remove unentrapped CF and contaminating SUVs, the vesicles were passed through an Econo-Pac lODG column (10 mL) and then a Sepharose2BCL column (1.6 X 60 cm), using 10 mM Pipes, 150 mM NaC1, and 1 mM EDTA, pH 7.0 (295 mOsM), as the eluting buffer. CF Release Assay. Release of CF from the vesicles at 25 “C was determined fluorometrically by monitoring the decrease in self-quenching (excitation 490 nm, emission 520 nm; Aminco SPF-5OOC spectrofluorometer) as the dye is released from the vesicles (Weinstein et al., 1984). Kinetic experiments were usually initiated by injecting 5-20 pL of peptide stock solution (0.37 mM mgn2a in buffer B) into a magnetically stirred, equilibrated cuvette containing 2 mL of 25-30 pM PS in buffer A. The mixing time was about 1 s. At the end of each run, the total releasable fluorescence wasdeterminedby theadditionof 20pLof lO%TritonX-lOO. Progress of the reaction is reported as the percentage of CF quenched, %Q, which is calculated from the formula [ 1 (F, F0)/(Fm Fo)] 100, where F, is the fluorescence at time t , Biochemistry, Vol. 31, No. 41, 1992 9913 FO is the base-line fluorescence of the vesicles, and F, is the fluorescence after Triton X-100 addition. As will be shown under Results, mgn2a induces the release of dye by an allor-none mechanism. Consequently, %Q equals the percentage of total dye remaining in the vesicles. The kinetics of dye release from SUVs were not significantly affected by the osmolarity of the bathing solution. This was determined by varying the osmolarity of buff