Antimicrobial Peptides Explained: How AMPs Kill Bacteria

Guides · Peptide science

Antimicrobial peptides are short, mostly positively charged molecules that almost every living thing makes to defend itself — and that your own body produces in your sweat, saliva, gut and white blood cells right now. They have been studied for fifty years as the answer to antibiotic resistance. The honest 2026 picture is more complicated, and more interesting, than that pitch.

What are antimicrobial peptides?

Antimicrobial peptides, also called host defence peptides or AMPs, are short peptides that kill or disable bacteria, fungi, viruses and parasites, usually by attacking the microbial membrane rather than a single enzyme. The Antimicrobial Peptide Database, the field’s main reference, listed 6,309 entries as of 1 January 2026 — 3,379 natural peptides, 2,290 synthetic ones and 373 predicted mainly by machine learning.

Two numbers from that database are worth holding onto, because they are measured rather than assumed. The average catalogued AMP is 28 residues long and carries a net charge of about +4.3. And 88% are cationic — which means roughly one in eight is neutral or, like human dermcidin, actually anionic. The textbook line that “AMPs are cationic” is a strong tendency, not a rule.

Structurally the database sorts them into four groups by three-dimensional shape: alpha-helical (magainin, LL-37), beta-sheet stabilised by disulfide bonds (the defensins), mixed alpha-beta (plant and beta-defensins), and non-alpha-beta extended peptides such as the tryptophan-rich indolicidin.

The human AMP arsenal

Of the catalogued peptides, 156 are human. They are not exotic — they are the reason a scrape on your skin usually does not become an infection.

The best studied is LL-37, the only cathelicidin humans make. It is stored as a 170-residue precursor called hCAP-18 in neutrophil granules and cut by the enzyme proteinase 3 to release the mature 37-residue peptide — which does, as its name suggests, begin with two leucines. On skin, further processing generates shorter fragments with their own activity profiles.

One regulatory detail is genuinely striking. The gene for LL-37, CAMP, is a direct target of the vitamin D receptor, and the vitamin D response element sits inside a primate-specific transposable element. Mice, rats and dogs do not have it. An entire arm of human innate immune regulation simply does not exist in the standard animal models — which is worth remembering before extrapolating rodent AMP data to people. (That the gene is induced is well established; whether vitamin D supplementation prevents infection in humans is a separate and much less settled question.)

How AMPs kill, and why they mostly spare your cells

The selectivity is not clever targeting. It is chemistry. Bacterial membranes present anionic phosphatidylglycerol and cardiolipin, wrapped in lipopolysaccharide or teichoic acids, and contain no sterols. Human cell membranes keep their anionic phospholipids on the inner leaflet, face the outside world with near-neutral phosphatidylcholine and sphingomyelin, and are stiffened by cholesterol. A cationic, amphipathic peptide is drawn to the first surface and largely ignores the second.

What happens after binding is less settled. The classic framework offers three models — barrel-stave, toroidal pore and carpet — but assignments are contested and often shift with peptide concentration and lipid composition. Membrane disruption is also not the whole story. Even the field’s canonical 2005 review noted that translocated peptides can inhibit cell wall synthesis, nucleic acid synthesis, protein synthesis or specific enzymes, and the database now classifies peptides by whether they act at the surface or on intracellular targets.

Many AMPs, LL-37 among them, also work as signalling molecules: recruiting neutrophils, inducing chemokines and modulating inflammation. For several of them the immune-modulating role may matter more in vivo than direct killing.

The resistance myth, and the counterexample that ends it

The standard argument is that because AMPs target a conserved physical property of the membrane rather than a single protein, escape would require expensive remodelling of a cell’s lipid chemistry, so resistance should evolve slowly. There is something to this: reported resistance typically raises the minimum inhibitory concentration two- to thirty-fold, against the hundred- to thousand-fold shifts seen with conventional antibiotics.

Why almost none of them became drugs

A short list of AMPs is genuinely in clinical use: gramicidin and polymyxin B in topical products since 1955, colistin since 1962, daptomycin since 2003, and nisin as an FDA-approved food preservative. That is close to the whole list — and no new antimicrobial peptide has been approved as an antimicrobial since daptomycin.

The failures are instructive. Pexiganan, a magainin analogue for diabetic foot infection, showed clinical comparability to oral ofloxacin in the 1990s — but the FDA held that equivalence to an active comparator was not adequate evidence, and the repeat programme of two Phase 3 trials in 389 patients missed superiority against vehicle plus standard wound care. Omiganan missed its primary endpoint. Iseganan’s ventilator-associated pneumonia trial was stopped early for futility. Murepavadin’s Phase 3 was halted after acute kidney injury appeared in 56% of treated patients versus 25 to 40% of controls.

Three structural problems sit behind most of this. Cost: solid-phase synthesis runs roughly $50 to $400 per gram of amino acid, against about $0.80 per gram for an aminoglycoside — a gap of two to three orders of magnitude that synthesis chemistry has not closed. Physiology: many AMPs lose activity at physiological salt concentrations, bind serum albumin, and are degraded by proteases. Toxicity: several are haemolytic at the concentrations required to kill bacteria, and increasing hydrophobicity to boost potency tends to increase haemolysis with it.

Where the field stands in 2026

The most visible current development is computational. A 2024 Cell paper mined 63,410 metagenomes and 87,920 microbial genomes to build AMPSphere, a catalogue of 863,498 non-redundant candidate peptides, most previously unknown; 100 were synthesised and showed activity against drug-resistant pathogens in vitro and in animals. A companion effort applied deep learning to extinct-organism proteomes. Whether any of this survives contact with clinical trials is the open question — discovery was never the bottleneck.

On the clinical side, PLG0206, a 24-residue engineered cationic peptide with orphan drug, QIDP and fast-track designations, entered a registrational Phase 2/3 trial in periprosthetic joint infection in March 2026, enrolling roughly 240 patients with treatment failure at 12 months as the primary endpoint. It is the most advanced AMP programme running.

Why AMPs are difficult to work with in the lab

This is where AMP research diverges most sharply from ordinary peptide handling, and where a lot of irreproducible data comes from.

Adsorption is the biggest and least appreciated problem. A 2015 study measuring mastoparan X, melittin and magainin 2 by HPLC found that at typical experimental concentrations 90% or more of the peptide can be lost from solution simply by sticking to the walls of glass vials and polypropylene tubes. Standard practice is to dilute in 0.01% acetic acid with 0.2% bovine serum albumin in low-binding tubes.

Salt sensitivity is equally concrete: LL-37 is active against MRSA, Proteus mirabilis and Candida albicans in low-salt media but not in media containing 100 mM sodium chloride — below the roughly 154 mM of physiological saline. Counter-ion matters too, since residual trifluoroacetate from synthesis and cleavage perturbs secondary structure and is itself cytotoxic; acetate and chloride salts of the same sequence have tested more potent. If you work with peptides at all, the general principles in our guides on solubility and diluents and reading an HPLC purity trace apply here with the volume turned up.

Frequently asked questions

Are antimicrobial peptides antibiotics?

Some are, in the regulatory sense — polymyxins, colistin and daptomycin are approved antimicrobial drugs and are structurally lipopeptides. Most are not drugs at all; they are components of innate immunity. Vancomycin, often lumped in, is a glycopeptide and not usually classed as a host defence peptide.

Can bacteria become resistant to antimicrobial peptides?

Yes. Resistance mechanisms are well documented, including surface charge modification, proteases and efflux, and colistin resistance has become mobile and transferable via the mcr-1 plasmid gene. Resistance appears to develop more slowly and with smaller shifts in potency than for conventional antibiotics, but it is not prevented.

Does vitamin D boost antimicrobial peptides?

Vitamin D directly induces transcription of the CAMP gene that encodes LL-37, and this is well established at the molecular level in human cells. Whether supplementation reduces infection risk in people is a separate clinical question that this molecular finding does not settle.

Why is LL-37 sold as a research peptide if it never became a drug?

Because it is one of the most-studied molecules in innate immunity, and research-grade material supports laboratory work on membrane biophysics, immune signalling and wound biology. That is a different thing from clinical validation: LL-37 is documented as haemolytic against human red blood cells, and it has never been approved for any human use.

Informational only — not medical advice · 21+. This article describes published laboratory and clinical research. It is not guidance for treating an infection, and nothing here is a recommendation to obtain or use any peptide.