Huey W. Huang

Magainin, found in the skin of Xenopus laevis, belongs to a broad class of antimicrobial peptides which kill bacteria by permeabilizing the cytoplasmic membrane but do not lyse eukaryotic cells. The 23-residue peptide has been shown to form an amphiphilic helix when associated with membranes. However, its molecular mechanism of action has been controversial. Oriented circular dichroism has detected helical magainin oriented perpendicular to the plane of the membrane at high peptide concentrations, but Raman, fluorescence, differential scanning calorimetry, and NMR all indicate that the peptide is associated with the head groups of the lipid bilayer. Here we show that neutron in-plane scattering detects pores formed by magainin 2 in membranes only when a substantial fraction of the peptide is oriented perpendicular to the membrane. The pores are almost twice as large as the alamethicin pores. On the basis of the in-plane scattering data, we propose a toroidal (or wormhole) model, which differs from the barrel-stave model of alamethicin in that the lipid bends back on itself like the inside of a torus. The bending requires a lateral expansion in the head group region of the bilayer. Magainin monomers play the role of fillers in the expansion region thereby stabilizing the pore. This molecular configuration is consistent with all published magainin data.

The argument and experimental evidence are presented for a two-state model that explains the action of both helical and beta-sheet antimicrobial peptides after they bind to the plasma membranes of cells. Each peptide has two distinct physical states of binding to lipid bilayers. At low peptide-to-lipid ratios (P/L), the peptide tends to adsorb in the lipid headgroup region in a functionally inactive state. At a P/L above a threshold value P/L, the peptide forms a multiple-pore state that is lethal to a cell. The susceptibility of a cell to an antimicrobial peptide depends on the value of P/L that is determined by the lipid composition of the cell membrane. This model provides plausible explanations for the experimental findings that the susceptibility of different bacteria to a peptide is not directly correlated to its binding affinity, different peptides preferentially kill different pathogens, and peptides exhibit varying levels of lytic activity against different eukaryotic cells.