How Antimicrobial Peptides Kill Bacteria

AMP Mechanisms

3,500+ AMPs cataloged

The Antimicrobial Peptide Database (APD) has cataloged over 3,500 natural antimicrobial peptides from across the tree of life, produced by organisms from bacteria to humans.

Wang et al., Nucleic Acids Research, 2026

Wang et al., Nucleic Acids Research, 2026

View as image

Antimicrobial peptides are ancient weapons. They predate antibiotics by hundreds of millions of years, present in organisms from insects to frogs to humans as a first line of defense against bacterial invasion. In 1987, Michael Zasloff isolated magainins from the skin of the African clawed frog Xenopus laevis, demonstrating that a 23-amino-acid peptide could kill bacteria across a broad spectrum without a specific enzymatic target.^[1] That discovery launched a field that now encompasses over 3,500 cataloged natural AMPs.^[2]

The central question since Zasloff's discovery has been: how do these peptides kill? The answer is not a single mechanism but a family of related strategies, all exploiting a fundamental difference between bacterial and mammalian cell membranes. This article covers the full landscape: the three classical models of membrane disruption, intracellular targeting, selectivity, and the therapeutic implications. For detailed coverage of the individual models, see The Barrel-Stave, Toroidal, and Carpet Models of AMP Action. For intracellular mechanisms, see Beyond Membrane Disruption: AMPs That Attack Intracellular Targets. For selectivity, see Why AMPs Target Bacterial Membranes but Spare Your Own Cells.

The electrostatic attraction: why AMPs find bacteria first

Before any pore forms, an AMP must reach the bacterial membrane and bind to it. This initial interaction is governed primarily by electrostatics.

Most AMPs carry a net positive charge of +2 to +9, conferred by lysine and arginine residues. Bacterial membranes are rich in negatively charged lipids: phosphatidylglycerol (PG), cardiolipin, and in Gram-negative bacteria, lipopolysaccharide (LPS) in the outer membrane. This creates a strong electrostatic attraction between the cationic peptide and the anionic bacterial surface.^[3]

Mammalian cell membranes present a different face. Their outer leaflet is dominated by zwitterionic (net neutral) lipids: phosphatidylcholine, sphingomyelin, and phosphatidylethanolamine. Cholesterol, which constitutes 20-25% of the mammalian membrane lipid content, further protects against AMP insertion by increasing membrane rigidity and condensing lipid packing. A 2026 biophysical study by Roldan et al. used multiple spectroscopic techniques to characterize how AMP-membrane interactions differ between bacterial and mammalian lipid compositions, confirming that electrostatic selectivity is the primary driver.^[4]

This charge-based selectivity is not absolute. At high concentrations, AMPs can damage mammalian cells (hemolysis), which is why therapeutic AMP design must optimize the balance between antimicrobial potency and host cell toxicity. For the full story on selectivity, see Why AMPs Target Bacterial Membranes but Spare Your Own Cells.

The barrel-stave model: the rarest mechanism

The barrel-stave model was the first proposed mechanism for AMP-mediated membrane disruption. In this model, peptide monomers insert perpendicularly into the lipid bilayer, oligomerize into a cylindrical arrangement (like staves of a barrel), and form a transmembrane pore. The hydrophobic faces of the peptides contact the lipid acyl chains, while the hydrophilic faces line the aqueous pore interior.^[5]

The barrel-stave model requires peptides with specific structural properties: they must be long enough to span the bilayer (approximately 22 amino acids in an alpha-helix to cross a 30-angstrom hydrophobic core), they must be sufficiently amphipathic to maintain both lipid and water contacts, and they must self-associate in the membrane.

Alamethicin, a 20-amino-acid peptide from the fungus Trichoderma viride, is the only well-established example of a barrel-stave pore former. Its pores are voltage-dependent, conductance-defined channels that can be studied by electrophysiology. The rarity of barrel-stave pores among natural AMPs suggests that this mechanism is structurally demanding: few peptides have the precise geometry required to form a stable, ordered transmembrane channel.

The toroidal pore model: bending the bilayer

The toroidal pore model (also called the wormhole model) accounts for the mechanism of most pore-forming AMPs, including magainins, melittin, and lacticin Q. Unlike barrel-stave pores, toroidal pores involve the lipid bilayer itself bending inward to form a continuous surface from the outer leaflet through the pore to the inner leaflet.^[5][6]

The process begins with peptides adsorbing onto the membrane surface in a parallel orientation. As local peptide concentration increases, the peptides impose strain on the outer leaflet, causing thinning of the membrane and expansion of the headgroup area. At a critical threshold (the peptide-to-lipid ratio), the bilayer curves inward, creating a pore lined by both peptide molecules and lipid headgroups. The result is a toroidal (donut-shaped) pore where the aqueous channel is bordered by an unbroken lipid monolayer surface.

Shai characterized this model extensively using pardaxin-derived peptides, demonstrating that the helix-hinge-helix structural motif common to many AMPs facilitates toroidal pore formation by allowing the peptide to bend with the curving bilayer.^[7] A 2002 review by Shai mapped the structural requirements for different AMP mechanisms, showing that peptide length, charge distribution, and amphipathicity determine whether a given AMP forms toroidal pores, barrel-stave pores, or acts through the carpet mechanism.^[6]

Toroidal pores are typically transient and dynamic, opening and closing as peptides exchange between the membrane surface and the pore structure. This distinguishes them from the stable, channel-like barrel-stave pores and helps explain why toroidal pore-forming AMPs often kill rapidly but without creating permanent membrane channels visible by electrophysiology.

The carpet model: overwhelming the membrane

The carpet model describes AMPs that kill by accumulating on the membrane surface at high density, like a carpet, until a threshold concentration destabilizes the bilayer entirely. No transmembrane pore forms. Instead, the membrane fragments into micelle-like structures, causing catastrophic lysis.^[5]

AMPs acting through the carpet model bind to the membrane surface with their hydrophobic faces oriented toward the lipid core but without inserting into the bilayer. As more peptides accumulate, they displace lipids, create curvature stress, and eventually the membrane loses structural integrity. The transition from intact bilayer to micelle debris is cooperative and concentration-dependent: below the critical threshold, the peptides sit harmlessly on the surface; above it, destruction is rapid.

Dermaseptins (from frog skin) and cecropins (from insect hemolymph) are examples of AMPs that primarily act through carpet-like mechanisms. The carpet model requires higher peptide concentrations than pore-forming models, which has implications for therapeutic dosing: carpet-acting AMPs may need higher tissue concentrations to achieve bactericidal effects.

For detailed comparisons of all three models with structural diagrams and energy landscapes, see The Barrel-Stave, Toroidal, and Carpet Models of AMP Action.

The aggregate model and other non-classical mechanisms

The three classical models do not cover every AMP. A fourth mechanism, the aggregate model, proposes that peptides form transient, unstructured aggregates within the membrane rather than organized pores or surface carpets. These aggregates create short-lived, irregularly sized channels that allow ion leakage and small molecule transit across the bilayer. The aggregate model may explain the activity of AMPs that show membrane disruption at concentrations too low for carpet-model lysis but whose pore properties do not match barrel-stave or toroidal geometry.

Other non-classical mechanisms include:

Lipid clustering and domain formation. Some AMPs preferentially bind to specific lipid types, sequestering them into domains and disrupting the normal lateral organization of the membrane. This can impair membrane-associated processes (respiration, cell division) without forming macroscopic pores.

Membrane thinning without pore formation. AMPs that insert shallowly into the outer leaflet increase the area of that leaflet relative to the inner leaflet, creating a thinning effect. At sufficient density, this thinning compromises the membrane's barrier function and allows leakage of ions and small molecules. Molecular dynamics simulations show that many AMPs thin the membrane substantially (by 3-5 angstroms) before reaching the concentrations needed for pore formation.

Oxidative damage. A subset of AMPs generate or potentiate reactive oxygen species (ROS) upon membrane interaction. This oxidative burst damages membrane lipids and proteins from within, accelerating cell death beyond what membrane disruption alone would achieve.

The boundaries between these mechanisms are fluid. Many AMPs likely use multiple mechanisms simultaneously or sequentially, depending on concentration, bacterial species, growth phase, and local membrane composition.

Synergy: AMPs working in combination

In biological systems, AMPs rarely act alone. The human innate immune system deploys multiple AMPs simultaneously (defensins, LL-37, lysozyme, lactoferricin), and these peptides often show synergistic killing, meaning their combined effect exceeds the sum of their individual activities.

Synergy between AMPs may arise from complementary mechanisms: one peptide disrupts the outer membrane (allowing access), while another targets the cytoplasmic membrane or intracellular targets. LL-37 and human beta-defensins, for example, show synergistic activity against both Gram-positive and Gram-negative bacteria in vitro. This combinatorial approach also reduces the effective concentration needed for each peptide, which reduces the risk of host cell toxicity.

The synergy principle extends to AMP-antibiotic combinations. Several studies have demonstrated that sub-inhibitory concentrations of AMPs can restore susceptibility of antibiotic-resistant bacteria to conventional drugs by permeabilizing the outer membrane and allowing the antibiotic to reach its intracellular target. This has led to interest in AMPs as antibiotic "adjuvants" rather than standalone therapeutics.

Beyond the membrane: intracellular AMP targets

A pivotal 2005 review by Brogden in Nature Reviews Microbiology challenged the membrane-centric view of AMP action.^[3] Brogden documented that many AMPs, after crossing or disrupting the membrane, interact with intracellular targets. These include:

DNA and RNA binding. Several cationic AMPs bind to bacterial DNA, inhibiting replication and transcription. Buforin II, a 21-amino-acid peptide from the Asian toad, penetrates the bacterial membrane without causing lysis and kills by binding to DNA and RNA inside the cell. Indolicidin, a tryptophan-rich bovine peptide, similarly crosses membranes at sub-lytic concentrations and inhibits DNA synthesis.

Protein synthesis inhibition. Proline-rich AMPs (such as oncocin and apidaecin from insects) enter bacterial cells through the inner membrane transporter SbmA and bind to the 70S ribosome, blocking translation. This mechanism is selective for bacteria because mammalian cells use 80S ribosomes with different binding pockets.

Cell wall synthesis disruption. Nisin, a lantibiotic produced by Lactococcus lactis, binds to lipid II, the essential precursor for peptidoglycan synthesis, blocking cell wall assembly. This dual mechanism (lipid II sequestration plus pore formation through lipid II-mediated oligomerization) makes nisin one of the most effective AMPs known.

Enzymatic inhibition. Some AMPs inhibit bacterial proteases, DNA gyrases, or metabolic enzymes. The mechanisms are often secondary to membrane effects but can contribute to killing at sub-lytic concentrations.

A 2026 study by Gao et al. showed that AMPs RI8 and RW8 target both bacterial membranes and genomic DNA, confirming the multi-target paradigm as a general feature rather than an exception.^[9] For a complete review of intracellular mechanisms, see Beyond Membrane Disruption: AMPs That Attack Intracellular Targets.

Speed of killing and resistance implications

AMPs kill fast. At bactericidal concentrations, many AMPs cause complete loss of membrane integrity within 2-4 minutes of contact. This speed has a critical consequence: it makes resistance development difficult.

Conventional antibiotics typically target a single enzyme or pathway (penicillin targets transpeptidase, fluoroquinolones target DNA gyrase). A single point mutation in the target can confer resistance. AMPs, by contrast, target the bacterial membrane itself, a fundamental structural feature that cannot be easily mutated away without compromising cell viability. Changing the entire lipid composition of the membrane is an energetically costly process that bacteria have limited capacity to achieve through single mutations.^[10]

That said, bacteria are not defenseless. Known AMP resistance mechanisms include:

• Membrane charge modification. Adding aminoarabinose to lipid A or lysylating phosphatidylglycerol reduces the net negative charge of the membrane, weakening the electrostatic attraction that AMPs depend on.
• Protease secretion. Some bacteria secrete outer membrane proteases that degrade AMPs before they reach the cytoplasmic membrane.
• Efflux pumps. ABC transporters can export AMPs that have entered the periplasmic space.
• Capsule and biofilm formation. Extracellular polysaccharides create a physical barrier that reduces AMP access to the membrane.

A 2026 study by Dubey et al. reviewed the current state of AMP resistance mechanisms, noting that while individual resistance mechanisms exist, the multi-target nature of AMPs means that comprehensive resistance requires simultaneous changes across multiple systems, which is rare in clinical settings.^[11]

For coverage of how AMPs compare to conventional antibiotics in addressing resistance, see Antimicrobial Peptides as Alternatives to Antibiotics: Can They Solve Resistance?

Structural features that determine mechanism

Not all AMPs use the same mechanism, and the structural features of a peptide predict which killing strategy it employs. A 2026 high-throughput screening study by Li et al. used mechanism-driven assays to classify AMPs by their membrane-targeting behavior, identifying structural determinants that push a peptide toward pore formation versus surface disruption.^[12]

The key structural parameters:

Amphipathicity. The segregation of hydrophobic and hydrophilic residues onto opposite faces of the helix determines how the peptide interacts with the bilayer interface. High amphipathicity favors membrane insertion and pore formation; moderate amphipathicity favors surface binding and carpet-like action.

Charge. Net positive charge of +2 to +4 is typical for membrane-active AMPs. Higher charges (+5 to +9) increase initial binding affinity but can reduce selectivity, causing hemolysis. Stephens et al. (2026) showed that selective serine substitutions in synthetic AMPs produced distinct mechanistic switches between Gram-negative killing pathways, demonstrating that even single-residue changes can shift the balance between pore formation and membrane solubilization.^[13]

Length. Peptides shorter than 12 amino acids generally cannot form stable transmembrane pores and instead act through carpet or intracellular mechanisms. Peptides of 20-40 amino acids can span the bilayer and are candidates for barrel-stave or toroidal pore formation.

Secondary structure. Alpha-helical AMPs (magainins, LL-37, melittin) are the most studied and most commonly form pores. Beta-sheet AMPs (defensins, protegrins) tend to form more rigid structures that can insert into membranes as pre-formed oligomers. Extended or loop structures (indolicidin, proline-rich peptides) are more likely to translocate without pore formation.

The therapeutic pipeline

The knowledge of AMP mechanisms has driven a growing therapeutic pipeline. Hancock and Sahl described in 2006 how host-defense peptides provide templates for two separate therapeutic strategies: direct antimicrobial killing and immunomodulation.^[10]

As of 2026, the APD database lists over 3,500 natural AMPs from six kingdoms of life.^[2] Murepavadin, which targets the outer membrane protein LptD of Pseudomonas aeruginosa, has advanced through Phase III clinical trials. Several AMPs are in clinical use as topical agents (nisin in food preservation, polymyxins for multidrug-resistant Gram-negative infections), and dozens more are in preclinical or early clinical development.

The challenge remains translating membrane-disrupting activity in vitro to systemic efficacy in vivo, where serum binding, protease degradation, and tissue distribution all reduce effective AMP concentrations. Strategies to address this include D-amino acid substitution, cyclization, lipidation, and encapsulation in nanoparticle delivery systems.

Open questions in AMP mechanism research

Despite decades of study, fundamental questions remain about how AMPs work in real biological contexts.

In vivo mechanism validation. Nearly all mechanistic studies use model membranes (liposomes, supported lipid bilayers) or simplified bacterial cultures. How AMP mechanisms operate in complex tissue environments, in the presence of serum proteins, mucus, and other host factors, is poorly understood. AMPs that form clear pores in vitro may act through entirely different mechanisms in a wound bed or mucosal surface.

Mechanism switching. Some AMPs appear to switch mechanisms depending on concentration, membrane composition, and bacterial species. Melittin, for instance, can form toroidal pores at low concentrations but acts through a carpet-like mechanism at high concentrations. Whether this plasticity is an exception or the rule for most AMPs is an active area of research.

Gram-positive versus Gram-negative differences. Most mechanistic models were developed using Gram-negative bacteria (which have an outer membrane containing LPS) or model membranes mimicking Gram-negative composition. Gram-positive bacteria lack an outer membrane but have a thick peptidoglycan layer that AMPs must traverse. The relative contribution of membrane disruption versus cell wall interactions in Gram-positive killing is less well characterized.

Biofilm penetration. Bacteria in biofilms are 100-1,000 times more resistant to AMPs than planktonic (free-floating) bacteria. Whether this resistance is primarily due to physical barriers (extracellular matrix), phenotypic changes (altered membrane composition), or both is critical for developing AMPs that can treat biofilm-associated infections.

For related coverage of AMPs from specific biological sources, see Amphibian Skin Peptides: The Pharmacy on a Frog's Back, Alpha-Defensins: The Neutrophil Peptides That Kill Bacteria on Contact, and Defensins in Your Lungs: The First Line of Airway Defense.

The Bottom Line

Antimicrobial peptides kill bacteria through multiple mechanisms, all rooted in their cationic, amphipathic structure and the electrostatic attraction to negatively charged bacterial membranes. The three classical membrane disruption models (barrel-stave, toroidal pore, and carpet) describe a spectrum from ordered transmembrane channels to catastrophic membrane dissolution. Many AMPs also act on intracellular targets including DNA, ribosomes, and cell wall synthesis machinery. This multi-target strategy explains both the broad-spectrum activity of AMPs and the difficulty bacteria face in developing comprehensive resistance. The over 3,500 natural AMPs cataloged to date represent a vast structural library for designing next-generation anti-infectives.

Frequently Asked Questions

How do antimicrobial peptides kill bacteria?

Antimicrobial peptides kill bacteria primarily by disrupting their cell membranes. Cationic AMPs are electrostatically attracted to the negatively charged bacterial membrane, where they insert and form pores (barrel-stave or toroidal models) or accumulate on the surface until the membrane fragments (carpet model). Many AMPs also cross the membrane and attack intracellular targets including DNA, RNA, ribosomes, and cell wall synthesis machinery. At bactericidal concentrations, membrane integrity is lost within 2-4 minutes.

What is the difference between barrel-stave and toroidal pore models?

In the barrel-stave model, peptides insert vertically into the membrane and oligomerize to form a channel lined entirely by peptides, like staves of a barrel. Only alamethicin is confirmed to use this mechanism. In the toroidal model, peptides cause the lipid bilayer to bend inward, creating a pore lined by both peptides and lipid headgroups. The toroidal pore is more common and accounts for the mechanism of magainins, melittin, and most other pore-forming AMPs.

Can bacteria become resistant to antimicrobial peptides?

Bacteria can develop partial resistance to AMPs through membrane charge modification (reducing electrostatic attraction), protease secretion (degrading peptides), efflux pumps (exporting peptides), and biofilm formation (blocking access). However, because AMPs target the membrane itself rather than a single enzyme, comprehensive resistance requires simultaneous changes across multiple systems. This makes full AMP resistance rare compared to resistance against conventional antibiotics.

Why don't antimicrobial peptides damage human cells?

AMPs preferentially target bacterial membranes because of electrostatic selectivity. Bacterial membranes are rich in negatively charged lipids (phosphatidylglycerol, cardiolipin, lipopolysaccharide), while mammalian membranes are dominated by zwitterionic (net neutral) lipids (phosphatidylcholine, sphingomyelin). Cholesterol in mammalian membranes further reduces AMP insertion by increasing membrane rigidity. This selectivity is concentration-dependent: at very high doses, AMPs can damage host cells.

How many antimicrobial peptides exist in nature?

The Antimicrobial Peptide Database (APD version 6, published 2026) catalogs over 3,500 natural AMPs from organisms across six kingdoms of life, including bacteria, fungi, plants, insects, amphibians, and mammals. The human body alone produces dozens of AMPs, including defensins, cathelicidins (LL-37), and histatins. Estimates suggest many more remain undiscovered, particularly in marine organisms and undersampled environments.

What is the carpet model of antimicrobial peptide action?

In the carpet model, AMPs adsorb onto the bacterial membrane surface in large numbers, covering it like a carpet. The peptides bind parallel to the membrane with their hydrophobic faces oriented toward the lipids but without inserting into the bilayer. When a critical concentration threshold is reached, the accumulated peptides destabilize the membrane, causing it to fragment into micelle-like structures. This results in catastrophic lysis. Dermaseptins and cecropins primarily use this mechanism.