Antimicrobial Peptides: The Innate Immune System's Own Antibiotics

Every multicellular organism that has ever been studied for it — plants, insects, amphibians, humans — makes its own antibiotics, and it made them hundreds of millions of years before a mold spore contaminated Alexander Fleming’s petri dish in 1928. These molecules are antimicrobial peptides (AMPs): short chains of amino acids, usually under 50 residues, that kill bacteria, fungi, and some enveloped viruses not by jamming a single metabolic enzyme the way most prescription antibiotics do, but by physically tearing open the microbial cell membrane. That mechanism is old, it’s broad-spectrum, and it’s remarkably hard for a microbe to evolve resistance against by mutating one gene — which is exactly why, as conventional antibiotic resistance climbs, the pharmaceutical industry keeps circling back to a molecule class the human innate immune system has been quietly relying on since before the immune system had T cells.

Two Human Families, One Membrane-Rupturing Idea

Humans make two major structural classes of antimicrobial peptide, and while they arrive at the same basic outcome — a dead microbial membrane — they get there with different molecular scaffolds.

Defensins are small, cysteine-rich peptides stabilized by three internal disulfide bonds that lock them into a compact, mostly beta-sheet fold. Humans produce six alpha-defensins: HNP1 through HNP4, stored in the azurophilic granules of neutrophils and released when the neutrophil degranulates to fight an infection, and HD5 and HD6, made by Paneth cells at the base of the intestinal crypts of Lieberkühn, where they help keep the gut’s bacterial population in check without depending on inflammation. A separate beta-defensin family (HBD1 through HBD4) is made mainly by epithelial cells — skin, lung lining, urinary tract — as a constant, low-level chemical barrier at every surface that touches the outside world. Both alpha- and beta-defensins share the same disulfide topology (Cys I–VI, II–IV, III–V) but differ in the spacing between cysteines and in exactly which cells express them.

Cathelicidins are structurally different: alpha-helical rather than beta-sheet, produced as an inactive precursor (hCAP18 in humans) that gets proteolytically cleaved to release the active peptide. Humans have only one cathelicidin gene, but its cleaved product, LL-37 (named for its first two residues, leucine-leucine, and its 37-residue length), is one of the best-studied AMPs in the literature — it’s made by neutrophils, macrophages, and epithelial cells, and it does far more than kill microbes directly, also acting as a chemoattractant that recruits other immune cells to a site of infection.

/posts/what-a-peptide-actually-is/ covers the general chemistry of how a peptide bond forms and how a chain folds; what matters here is that both defensins and cathelicidins fold into a specific geometric trick — most of the positively charged residues cluster on one face of the molecule and most of the hydrophobic residues cluster on the other. That amphipathic arrangement, not any single active site, is the entire mechanism.

How They Actually Kill: Physics, Not a Lock and Key

Most antibiotics work the way most drugs work: a specific molecule binds a specific target — a ribosomal subunit, a cell-wall synthesis enzyme, a DNA gyrase — and blocks it. Antimicrobial peptides mostly don’t do that. Instead, their cationic, amphipathic structure lets them insert directly into a lipid bilayer and physically disrupt it, a mechanism that’s been described through three overlapping structural models depending on the specific peptide and membrane involved.

BARREL-STAVE MODEL              TOROIDAL PORE MODEL           CARPET MODEL
(peptides line a channel)       (peptides + lipids co-form     (peptide coats surface,
                                  the pore rim)                  then membrane collapses)

  membrane surface                membrane surface               peptides adsorb flat
  ||||||||||||||||                |||||\   /|||||                ||||||||||||||||
  |||[P]|[P]||||||   <- peptides  ||||[P]-[P]||||   <- lipid      [P][P][P][P][P][P]
  |||[P]| |[P]|||||     line a    |||headgroups curve             (accumulate until
  |||[P]|_|[P]|||||     channel   |||into the pore wall           a critical density
  ||||||||||||||||      wall      |||||/   \|||||                 is reached)
  bilayer intact                  bilayer bent, mixed              |
  except for the pore             peptide-lipid pore               v
                                                                 membrane disintegrates
                                                                 like a detergent action

LL-37 in particular has been shown to behave differently depending on membrane composition — sometimes forming oligomeric, fibril-like structures that assemble into pores resembling a distorted toroidal model, and sometimes acting through a carpet-type mechanism where the peptide lies flat along the membrane surface until enough peptide has accumulated to destabilize the bilayer wholesale, the same basic physics behind how a detergent dissolves a lipid membrane. In every model, the outcome is the same: the transmembrane electrochemical potential collapses, the cell’s contents leak out, and the microbe dies within minutes rather than the hours a metabolism-blocking antibiotic typically takes.

Because it’s the physical property of the lipid bilayer being exploited — not a specific protein target — this class of mechanism is dramatically harder for bacteria to develop single-point resistance mutations against compared to a conventional antibiotic that can be defeated by one mutated enzyme active site.

Selectivity: Why AMPs Don’t Shred Your Own Cells

The obvious question is why a molecule that dissolves lipid membranes on contact doesn’t kill the host’s own cells first. The answer is electrostatics, and it comes down to a real compositional difference between microbial and mammalian membranes.

Bacterial membranes are rich in anionic phospholipids — phosphatidylglycerol and cardiolipin carry a net negative charge on their headgroups, and Gram-negative bacteria add another negatively charged layer on top in the form of lipopolysaccharide (LPS) in the outer membrane. Mammalian plasma membranes, by contrast, are built mostly from zwitterionic phospholipids like phosphatidylcholine and phosphatidylethanolamine, which carry no net charge, and are further stabilized by cholesterol, which most bacterial membranes lack entirely. Since defensins and cathelicidins carry a strong net positive charge — LL-37 alone has a net charge of roughly +6 at physiological pH — they’re electrostatically drawn toward the anionic bacterial surface far more strongly than toward the comparatively neutral mammalian one, and cholesterol in mammalian membranes further resists the kind of curvature and insertion AMPs depend on.

This selectivity is real but not absolute — it’s a difference of degree, driven by relative charge density, not an airtight biochemical firewall. At high enough concentrations, most AMPs will also damage mammalian membranes, which is precisely the toxicity ceiling that has complicated turning several of them into systemic drugs.

Already in the Clinic: Polymyxins as Proof the Mechanism Works

The idea of a membrane-disrupting cationic peptide as an antibiotic isn’t hypothetical — it’s already standard practice in hospital medicine, just not usually via a human-derived molecule. Colistin (polymyxin E) and polymyxin B are bacterially-derived cyclic lipopeptides — a cationic peptide ring of seven amino acids attached to a hydrophobic fatty-acid tail — that kill Gram-negative bacteria through a mechanism described by the Shai-Matsuzaki-Huang model: electrostatic attraction to the negatively charged lipopolysaccharide in the outer membrane, followed by insertion and permeabilization that lets cell contents leak out, functionally the same carpet/toroidal logic that defensins and LL-37 use against their targets.

Because polymyxins remain active against many carbapenem-resistant and multidrug-resistant Gram-negative organisms, they’ve become a genuine last-resort therapy — but the same nonspecific membrane-targeting mechanism that makes them broad-spectrum also makes them dose-limiting toxic. Reported nephrotoxicity rates for polymyxin therapy range from roughly 15% to 60% of treated patients, driven by colistin accumulating in renal proximal tubular cells and triggering endoplasmic reticulum stress, mitochondrial dysfunction, and oxidative injury in those cells. Polymyxins are the clearest existing proof that a membrane-disrupting peptide can be a real, prescribable antibiotic — and simultaneously the clearest warning about the toxicity trade-off that comes bundled with the mechanism.

Computing the Trait That Makes It Work: Hydrophobic Moment

The amphipathicity that makes an AMP selective for microbial membranes is something you can actually calculate directly from a peptide’s sequence, using a metric called the hydrophobic moment — a measure of how asymmetrically hydrophobicity is distributed around a helical wheel. It’s a real, standard bioinformatics calculation used in AMP design and screening pipelines, not just a qualitative description:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

import cmath

# Eisenberg hydrophobicity scale (partial, common residues)
HYDROPHOBICITY = {
    'A': 0.62, 'R': -2.53, 'N': -0.78, 'D': -0.90, 'C': 0.29,
    'Q': -0.85, 'E': -0.74, 'G': 0.48, 'H': -0.40, 'I': 1.38,
    'L': 1.06, 'K': -1.50, 'M': 0.64, 'F': 1.19, 'P': 0.12,
    'S': -0.18, 'T': -0.05, 'W': 0.81, 'Y': 0.26, 'V': 1.08,
}

def hydrophobic_moment(sequence, delta_degrees=100):
    """Mean hydrophobic moment for an alpha-helical peptide.
    delta_degrees=100 is the standard per-residue turn angle
    for an alpha helix (3.6 residues per turn)."""
    total = complex(0, 0)
    for i, residue in enumerate(sequence):
        angle = cmath.pi * i * delta_degrees / 180
        h = HYDROPHOBICITY.get(residue, 0.0)
        total += h * cmath.exp(complex(0, angle))
    return abs(total) / len(sequence)

# LL-37, first 18 residues (the helical core region)
ll37_core = "LLGDFFRKSKEKIGKEFK"
print(f"{hydrophobic_moment(ll37_core):.3f}")

Run against a genuinely amphipathic helical AMP sequence, this returns a high mean hydrophobic moment — a strong asymmetry between the peptide’s hydrophobic and hydrophilic faces. Run against a random or non-amphipathic sequence of the same amino acid composition, the value collapses toward zero, because the same hydrophobic and charged residues are scattered around the helix rather than segregated onto opposite faces. This single calculable number is a large part of why some peptide sequences are potent membrane disruptors and others, built from nearly the same amino acid palette, aren’t.

The Drug-Development Graveyard

If the mechanism is this robust and this hard to evolve resistance against, the obvious question is why AMP-based drugs aren’t already a bigger part of the antibiotic arsenal beyond the polymyxins. The honest answer is that dozens of AMP-derived drug candidates have entered clinical trials over the past three decades, and the great majority have failed — not because the underlying biology is wrong, but because the properties that make a good innate-immune weapon don’t automatically make a good systemic drug.

Pexiganan, a synthetic analog of the frog-derived peptide magainin II, is the textbook case. It showed broad-spectrum activity, including against MRSA, and was developed as a topical treatment for infected diabetic foot ulcers. It cleared safety testing but failed to gain FDA approval after Phase III trials showed no statistically meaningful efficacy advantage over existing conventional antibiotic treatment for the same indication — not a safety failure, but a failure to clear the bar of “better than what already exists,” which is a much higher bar for a novel, expensive-to-manufacture peptide than for a cheap generic antibiotic.

None of these are arguments that the mechanism doesn’t work — polymyxins prove it does. They’re arguments that turning a peptide the immune system evolved to use locally, in short bursts, at a wound or mucosal surface into a systemically dosed, mass-manufactured, shelf-stable pharmaceutical is a substantially harder engineering problem than the underlying microbiology would suggest by itself.

Honest Trade-offs

• The membrane-rupture mechanism resists single-point resistance mutations, but it isn’t resistance-proof. Some bacteria — certain Staphylococcus aureus and Salmonella strains among them — have evolved membrane modifications (like reducing net negative surface charge) that measurably reduce AMP susceptibility, so “hard to evolve resistance against” is a relative advantage over conventional antibiotics, not an absolute immunity claim.
• Selectivity for microbial over mammalian membranes is a matter of degree, not an on/off switch. At high enough local concentrations, most AMPs will also damage host cells, which is exactly the mechanism behind the significant nephrotoxicity rates seen with polymyxin therapy in the clinic.
• The same short length and instability that let AMPs act fast in the body also make them pharmaceutically inconvenient. Rapid proteolytic degradation is a feature for a peptide meant to act locally and disappear, and a liability for a peptide a company wants to formulate into a stable, shelf-stable, systemically dosed drug.
• Clinical trial failures for AMP candidates have mostly been efficacy failures relative to existing drugs, not safety failures. That’s a genuinely different, and in some ways harder, bar to clear — a novel, expensive-to-manufacture peptide has to outperform a cheap, well-understood generic antibiotic, not just prove it’s non-toxic.
• Topical and last-resort systemic use cases remain the realistic near-term niche, not broad first-line replacement of conventional antibiotics. Polymyxins are used specifically because other options have failed, and most AMP drug candidates in active development are aimed at topical, localized, or genuinely resistant infections rather than competing head-on with first-line oral antibiotics.

Verdict

Antimicrobial peptides are proof that the most robust antibiotic mechanism available isn’t a clever new one — it’s the oldest one, older than adaptive immunity itself, and it works by physics rather than by finding a single molecular lock to pick. Defensins and cathelicidins show the human body has been running this membrane-disruption strategy locally and continuously for as long as there’s been an innate immune system to run it, and the polymyxin antibiotics already in clinical use prove the same basic mechanism translates into a real, prescribable drug when the toxicity trade-off is managed carefully. What the drug-development track record actually shows is not that the biology is flawed, but that a molecule optimized by evolution for rapid, local, disposable action at a wound site or mucosal surface has to be re-engineered substantially — for stability, for cost, for a wider therapeutic margin — before it can compete as a shelf-stable, systemically dosed pharmaceutical against cheap, well-understood generic antibiotics. As resistance to those generics keeps climbing, that re-engineering problem is exactly where the next generation of antibiotic development money is going.

Sources

• Frontiers — The Human Antimicrobial Peptides Dermcidin and LL-37 Show Novel Distinct Pathways in Membrane Interactions
• PMC — Molecular Dynamics Simulations of Human Antimicrobial Peptide LL-37 in Model POPC and POPG Lipid Bilayers
• PMC — Structures of Human Host Defense Cathelicidin LL-37 and Its Smallest Antimicrobial Peptide KR-12 in Lipid Micelles
• PMC — Human α-Defensin 6: a Small Peptide That Self-Assembles and Protects the Host by Entangling Microbes
• PMC — Crystal Structures of Human α-Defensins HNP4, HD5, and HD6
• eLife — Colistin Kills Bacteria by Targeting Lipopolysaccharide in the Cytoplasmic Membrane
• PMC — Molecular Mechanisms of Colistin-Induced Nephrotoxicity
• PMC — Comparison of Nephrotoxicity of Colistin with Polymyxin B Administered in Currently Recommended Doses
• Frontiers — Review: Lessons Learned From Clinical Trials Using Antimicrobial Peptides (AMPs)
• Frontiers — Unlocking the Power of Antimicrobial Peptides: Advances in Production, Optimization, and Therapeutics