Antimicrobial Peptides: What the Science Actually Shows

Antimicrobial peptides (AMPs) are short chains of amino acids, typically between 10 and 50 residues, that are encoded directly in the genome of an organism and produced as part of its innate immune defense. They are found in essentially every branch of life that has been examined, from bacteria and plants to insects, amphibians and humans. Because many of them also talk to the immune system rather than simply killing microbes, researchers increasingly call them host defense peptides, a term that captures the dual role more honestly.

Two physicochemical features define most of the class. The first is a net positive charge, usually between +2 and +9, carried by lysine and arginine residues. The second is amphipathicity: when the peptide folds, hydrophobic side chains cluster on one face and charged side chains on the other. That architecture lets the peptide sit at the interface between water and a lipid membrane, which is where most of its biology happens. Structurally, AMPs fall into a few recurring shapes: alpha-helices such as LL-37 and the magainins, disulfide-stabilized beta-sheets such as the defensins, extended peptides rich in proline or tryptophan such as indolicidin, and small loop structures.

The scale of the class is substantial. The Antimicrobial Peptide Database APD3, in its 2016 release, catalogued 2,619 entries, including 321 peptides from plants, 261 bacteriocins from bacteria and 1,972 animal host defense peptides. The database has continued to grow since, and computational design now adds candidate sequences faster than laboratories can test them.

It helps to be precise about what AMPs are not. Most conventional antibiotics are small molecules produced by soil microorganisms that inhibit a specific bacterial enzyme or ribosomal function. AMPs are polymers built by the host's own ribosomes (or, for some microbial peptides, by non-ribosomal enzyme assembly lines), and they usually act on a physical structure, the membrane, rather than on a single protein target. If the general vocabulary of peptide biology is new to you, the overview in how peptides work covers the basics of peptide structure, signaling and stability that the rest of this article assumes.

In humans, two families dominate. The defensins are cationic peptides of roughly 30 amino acids containing six conserved cysteines whose three disulfide bridges lock them into a mostly beta-sheet fold. Humans express six alpha-defensins: HNP1 to HNP4, stored in neutrophil granules, plus HD5 and HD6, secreted by Paneth cells in the small intestine. The beta-defensins, hBD1 to hBD4 being the best characterized, are produced mainly by epithelial surfaces: skin, airway, gut and urogenital tract. Lehrer and Lu's review of alpha-defensins in human innate immunity describes this division between myeloid and enteric peptides and the breadth of their antimicrobial and antiviral activity.

The second family is the cathelicidins. Humans have exactly one, stored as the precursor hCAP18 and cleaved to release the active 37-residue peptide beginning with two leucines, hence the name LL-37. It is produced by neutrophils, macrophages, keratinocytes and epithelial cells, and it is the single most studied human AMP. Its sequence, LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES, folds into an amphipathic helix in the presence of membranes. A detailed monograph is available in the LL-37 guide.

Saliva contributes a third group: the histatins, histidine-rich peptides of which histatin 5 is the most potent antifungal. Rather than punching holes in the fungal membrane, histatin 5 binds cell wall proteins of Candida albicans, enters through polyamine transporters in an energy-dependent way, and kills from the inside. Other human AMPs include dermcidin in sweat, lactoferricin derived from lactoferrin in milk and secretions, and the bactericidal/permeability-increasing protein in neutrophils.

Outside humans, the catalogue is wide. Amphibian skin yields the magainins from Xenopus laevis, plus temporins and dermaseptins. Insects produce cecropins, drosocin and defensin-like peptides. Venoms are a rich source: melittin from honeybee venom and mastoparan from wasp venom are powerful membrane-active peptides, though their lack of selectivity between microbial and human cells is exactly why they remain laboratory tools rather than drug candidates. Plants make thionins, plant defensins and cyclotides, which are stabilized by a cyclic backbone and a cystine knot. Bacteria themselves produce bacteriocins, including nisin, which is used as a food preservative and acts by binding the cell wall precursor lipid II.

Selectivity comes first. Bacterial membranes are rich in anionic phospholipids such as phosphatidylglycerol and cardiolipin, and their surfaces carry lipopolysaccharide in Gram-negative species or teichoic acids in Gram-positive species. Healthy mammalian cell membranes present a largely neutral outer leaflet of phosphatidylcholine and sphingomyelin, stiffened by cholesterol. A cationic peptide is therefore electrostatically steered toward the microbe. This is a preference, not an absolute rule, and the margin between the concentration that kills bacteria and the concentration that damages human cells is the central pharmacological problem of the field.

Once bound, membrane-active peptides accumulate until a threshold is crossed. Several models describe what follows: the barrel-stave model, in which peptides line a pore like staves of a barrel; the toroidal pore model, in which the lipid monolayer bends through the pore alongside the peptides; and the carpet model, in which peptides coat the surface and solubilize it in a detergent-like fashion. The practical consequence is the same in each case: loss of the transmembrane potential, leakage of ions and metabolites, and rapid death, often within minutes. The 2025 review by Oliveira Júnior and colleagues in Nature Reviews Microbiology surveys this structural and mechanistic diversity in detail.

Not all AMPs kill by lysis. Some cross the membrane with little damage and attack intracellular targets: binding DNA and RNA, inhibiting protein synthesis, or blocking chaperone function. Histatin 5 is the clearest example of a non-lytic peptide. Pore formation has largely been disproven for it, and the current picture, as reviewed by Puri and Edgerton, is of intracellular targets, mitochondrial dysfunction and oxidative stress, with cell death ultimately driven by ion imbalance and osmotic volume dysregulation. Other peptides target biosynthetic machinery: nisin sequesters lipid II, and the peptidomimetic murepavadin binds LptD, the protein that inserts lipopolysaccharide into the outer membrane of Pseudomonas aeruginosa.

The approved lipopeptide daptomycin illustrates how specific the chemistry can get. Its activity strictly requires calcium ions. Huang's 2020 analysis in Biochimica et Biophysica Acta Biomembranes describes the formation of a defined daptomycin-calcium-phosphatidylglycerol complex that behaves as a transient ionophore, depolarizing the membrane through ion leakage rather than through a classical stable pore. Separate work also implicates interference with cell wall precursor lipids. Mechanism in this field is rarely a single story.

Yes, and for several human peptides the immune signaling role may matter more in vivo than direct killing. At the concentrations actually reached in tissue and in the presence of physiological salt and serum proteins, the direct bactericidal activity of LL-37 is often modest. What remains robust is its signaling activity.

LL-37 functions as an alarmin. It is chemotactic for neutrophils, monocytes and T cells, largely through the formyl peptide receptor 2, it neutralizes lipopolysaccharide and so dampens endotoxin-driven inflammation, it promotes angiogenesis and re-epithelialization in wound models, and it influences apoptosis and cell differentiation. Fabisiak and colleagues summarized this pleiotropic profile in Pharmacological Reports in 2016, noting activities that run in both pro-inflammatory and anti-inflammatory directions depending on context. Defensins behave similarly: human beta-defensin 2 is chemotactic through CCR6, linking epithelial innate defense to adaptive immunity.

The same versatility carries a warning. LL-37 forms complexes with self DNA and RNA that trigger plasmacytoid dendritic cells through endosomal Toll-like receptors, a mechanism implicated in the pathology of psoriasis. Abnormal cathelicidin processing is also associated with rosacea. A molecule that amplifies immune signaling is not automatically beneficial, and more is not better.

This immunomodulatory branch is where the antimicrobial peptide literature overlaps with the broader interest in peptides that tune inflammation rather than attack pathogens. The tripeptide KPV, an alpha-MSH fragment studied for its anti-inflammatory signaling, is covered in the KPV guide, and thymosin alpha-1, an immune-modulating peptide approved in a number of countries outside the United States, is discussed separately. A comparative look at the evidence base for these molecules is available in our review of peptides studied for inflammation. None of these are antibiotics, and none should be thought of as a substitute for one.

The appeal is straightforward. Antimicrobial resistance is one of the most serious problems in infectious disease, the conventional antibiotic pipeline has been thin for decades, and AMPs offer a mode of action that differs fundamentally from beta-lactams, fluoroquinolones or macrolides. Several properties are genuinely attractive: killing is fast, which leaves little time for adaptive responses; the target is a membrane architecture rather than a single mutable protein; many AMPs retain activity against metabolically dormant persister cells and against biofilms, where conventional antibiotics often fail; and synergy with existing antibiotics has been demonstrated repeatedly in vitro, including permeabilization of the Gram-negative outer membrane to let other drugs in.

The frequently repeated claim that bacteria cannot evolve resistance to AMPs is, however, wrong. Joo, Fu and Otto catalogued bacterial resistance strategies in Philosophical Transactions of the Royal Society B in 2016, and the list is long. Bacteria reduce the net negative charge of their surface by adding L-lysine to phosphatidylglycerol through MprF, by D-alanylating teichoic acids, or by decorating lipid A with 4-amino-4-deoxy-L-arabinose and phosphoethanolamine. They secrete proteases that degrade peptides, deploy efflux pumps, shield themselves with capsules and biofilm matrix, and sense host peptides through two-component systems such as PhoPQ and GraRS in order to switch these defenses on.

Clinical experience confirms the point. Resistance to colistin, a peptide antibiotic held in reserve for multidrug-resistant Gram-negative infections, arises through chromosomal mutations and, more worryingly, through the plasmid-borne mcr gene family, which is horizontally transferable between bacteria and species. A resistance mechanism that can move on a plasmid is exactly what makes a last-line agent fragile. The review by Moravej and colleagues in Microbial Drug Resistance covers these mechanisms alongside peptide features and action.

There is a further argument that deserves attention. Because therapeutic AMPs often resemble the peptides of human innate immunity, selecting for resistance to a drug version could in principle select for reduced susceptibility to our own defensins and cathelicidins. The magnitude of that risk is debated and hard to quantify, but it is one reason some researchers favor peptidomimetics and engineered sequences with no close human homolog.

Peptide antibiotics are not hypothetical. Several have been in clinical use for decades. What they have in common is instructive: nearly all are microbial products made by non-ribosomal peptide synthetases, not human host defense peptides.

The polymyxins, polymyxin B and polymyxin E (colistin), are cyclic lipopeptides from Paenibacillus polymyxa. They were introduced in the late 1950s, then largely abandoned in the 1970s because of toxicity once safer agents appeared, and were revived in the 2000s when carbapenem-resistant Gram-negative infections left few alternatives. They bind lipid A in the outer membrane of Gram-negative bacteria and disorganize it. Their limiting problem is dose-related nephrotoxicity, with neurotoxicity less frequent. Kelesidis and Falagas, reviewing polymyxin safety in Expert Opinion on Drug Safety, describe a narrow therapeutic range in which kidney injury is common and relates both to host factors and to drug exposure, though toxicity is usually reversible.

Daptomycin was approved by the US Food and Drug Administration in September 2003 for complicated skin and skin structure infections, with later approval for Staphylococcus aureus bloodstream infection including right-sided endocarditis. It is active against Gram-positive organisms including MRSA and vancomycin-resistant enterococci. It is inactivated by pulmonary surfactant, so it is not used for pneumonia, a reminder that a membrane-active mechanism is only as useful as the environment allows.

The pattern is clear: the approved peptide antibiotics are old microbial molecules, and the human peptides that generate the most scientific excitement have produced no approved drug. These prescription medicines are used in supervised hospital settings. Nothing in this article constitutes dosing or administration guidance.

Reviews of the field generally count on the order of twenty peptide candidates with antibacterial or antifungal activity moving through clinical phases at any given time, with most of them in topical or local applications rather than systemic therapy. The history of the late-stage failures is the most informative part of the story.

Pexiganan (MSI-78, marketed as Locilex) is a synthetic analog of magainin 2 from frog skin, formulated as a topical cream for mildly infected diabetic foot ulcers. It was tested in two phase 3 trials, OneStep-1 and OneStep-2, registered on ClinicalTrials.gov, which compared the cream plus standardized wound care against vehicle plus standardized wound care. The program did not meet its primary endpoint of superiority over vehicle. This was the second time pexiganan had failed to convince regulators, an earlier program in the 1990s having also fallen short.

Iseganan, a protegrin-1 analog developed as an oral rinse for mucositis, likewise failed to demonstrate benefit in phase 3. Omiganan, derived from bovine indolicidin, has been studied topically for catheter-site infections and inflammatory skin conditions with mixed results. Murepavadin, a peptidomimetic that targets LptD in Pseudomonas aeruginosa and represented one of the most mechanistically interesting candidates in years, had its intravenous phase 3 program in nosocomial pneumonia halted in 2019 after an unexpected rate of acute kidney injury. Inhaled formulations have since been explored to limit systemic exposure.

Other approaches remain active: brilacidin and related defensin mimetics built on non-peptide scaffolds, lipopeptide derivatives of polymyxin engineered for lower kidney toxicity, bacteriocin-based products, and a growing number of sequences designed by machine learning models and then synthesized and screened. LL-37 itself has been examined in small early-phase topical studies, including in venous leg ulcers, which is a long distance from a licensed systemic antibiotic.

Every peptide named in this section is investigational. None of them is approved for the uses described, and discussing a trial is not an endorsement of self-experimentation. Anyone with a suspected infection needs assessment by a clinician, not a peptide.

Selectivity and toxicity. AMPs act mainly on membranes, and their discrimination between microbial and human membranes is relative rather than absolute. Many peptides that are impressive in a microbiology plate show hemolytic or cytotoxic effects at concentrations close to their antimicrobial range. The therapeutic window can be narrow, and in systemic use it has repeatedly proven too narrow, as the polymyxin and murepavadin experiences show.

Stability. Peptides made of L-amino acids are substrates for the proteases of both the host and the pathogen. Circulating half-lives are typically measured in minutes, and bacterial proteases such as staphylococcal aureolysin degrade host peptides as a deliberate defense. Oral administration is generally impossible for unmodified peptides because of gastric and intestinal degradation.

Loss of activity under physiological conditions. This is the most underappreciated obstacle. AMP binding depends on electrostatics, so physiological salt concentrations screen the charge interaction and can substantially reduce killing. Serum proteins bind peptides, divalent cations compete for anionic sites on bacterial surfaces, and the acidic or protein-rich environment of an infected tissue further changes behavior. Activity measured in dilute buffer routinely overestimates activity in a patient.

Cost and manufacturing. Solid-phase synthesis of a 30-residue peptide at pharmaceutical scale is expensive compared with fermentation or small-molecule chemistry, and cost of goods matters a great deal for an anti-infective that must compete with generic antibiotics. Recombinant expression is complicated by the fact that the product is toxic to the expression host.

The field's responses are well defined, even if none has yet produced an approved host defense peptide drug: substitution with D-amino acids or unnatural residues to resist proteases, backbone cyclization, lipidation to improve membrane partitioning, truncation to short active cores, peptidomimetic scaffolds that keep the charge and amphipathicity while discarding the peptide bond, encapsulation in nanoparticles or hydrogels for local delivery, and deep learning models that explore sequence space far faster than medicinal chemistry alone. Topical and localized routes remain the most realistic near-term applications because they sidestep systemic toxicity and serum inactivation at once.

Several peptides discussed in this article, including LL-37 and various defensin fragments, are offered by chemical suppliers labelled for laboratory research use only. It is worth being explicit about what that label means and does not mean. It means the material has not been evaluated or approved by the FDA, the EMA or any comparable authority for use in people. It does not mean the peptide is a medicine with a different sales channel, and it carries no assurance of identity, purity, endotoxin content or sterility.

There is also a conceptual confusion worth clearing up. Within the peptide research community, the molecules most often discussed are immunomodulators and signaling peptides rather than antibiotics. KPV and thymosin alpha-1 are studied for their effects on inflammatory and immune signaling, not for killing bacteria. LL-37 genuinely sits in both categories, which is precisely why its risk profile is complicated: the same molecule that recruits neutrophils and supports wound repair also forms immunostimulatory complexes with self nucleic acids and is implicated in inflammatory skin disease. For a broader comparison of the evidence behind immune-modulating peptides, see our overview of peptides studied for inflammation.

Legal status varies by jurisdiction. In most of the United States and the European Union, research peptides are not authorized for human administration, importation rules differ, and some peptides fall under anti-doping prohibitions in sport. Compounded or imported preparations have been the subject of regulatory warning letters.

Medical disclaimer. This article is for educational purposes only. It does not provide dosing, administration or usage protocols of any kind, and it is not medical advice. Antimicrobial peptides that are approved as medicines, such as colistin and daptomycin, are hospital-reserved prescription drugs that require clinical monitoring. Host defense peptides such as LL-37 are not approved for human use, and preclinical or animal data should not be read as evidence of human benefit. Suspected infection requires prompt evaluation by a qualified healthcare professional, and no peptide should be used in place of appropriate antibiotic care.