Chapter 25: Antimicrobial Peptides — The Natural Weapons That Could Solve Antibiotic Resistance

"It is not difficult to make microbes resistant to penicillin in the laboratory by exposing them to concentrations not sufficient to kill them, and the same thing has occasionally happened in the body." — Alexander Fleming, Nobel Lecture, December 11, 1945

Overview

A frog sits in a tank of unsterilized water with a fresh surgical incision in its abdomen. The wound does not become infected. In 1987 a scientist at the National Institutes of Health noticed that this was strange, went looking for the reason, and found two short peptides in the frog's skin that killed bacteria on contact. He named them magainins, from the Hebrew word for shield.

That observation launched a field. Nearly four decades later, the field has produced an enormous amount of excellent science, several thousand cataloged peptide sequences, a genuine and well-characterized mechanism of action, and — as of this writing — not one broad-spectrum systemic antibiotic derived from that work.

This chapter is about why. It is the cleanest case in the book of science that is real, obstacles that are real, and a verdict that is genuinely unavailable. If you have been reading confident articles telling you that antimicrobial peptides are about to solve antibiotic resistance, or equally confident articles telling you they are a decades-long dead end, you have been reading two different kinds of overclaim about the same body of work.

The chapter also contains a correction that most popular coverage gets backwards, worth flagging now so you read looking for it. Peptide antibiotics are not hypothetical. They are in hospital pharmacies today and have been for sixty years. Colistin, daptomycin, vancomycin, bacitracin, gramicidin, polymyxin B — every one a peptide, every one approved, several of them the drugs a physician reaches for when nothing else works. The sentence "antimicrobial peptides don't work as drugs yet" is simply false, and the accurate statement is narrower and much more interesting.

You will also learn something structural that applies well beyond this chapter: the main barrier to new antibiotics is not, at this point, primarily scientific. It is economic, and it is the kind of economic problem no laboratory result can fix. Holding both facts at once — the biology is promising, the market is broken — is the intellectual work of this chapter.

In this chapter, you will learn to:

• State the resistance burden accurately, including which of the two commonly cited figures you mean
• Explain what antimicrobial peptides are, how old they are, and why nearly every multicellular organism makes them
• Describe membrane disruption and say precisely why it makes resistance harder rather than impossible
• Recognize the major AMP families — defensins, cathelicidins, magainins — and what each taught the field
• Define hemolysis and the selectivity index, and explain why selectivity is the obstacle that has killed the most promising candidates
• Explain the economics of antibiotics well enough to see why they are as much of a barrier as biology
• Distinguish the evidence state of topical AMPs from systemic AMPs
• Name the peptide antibiotics already in clinical use and explain what their existence proves and does not prove
• Write a 🔬 verdict that is honest rather than evasive, by naming what would resolve it

Learning Paths

All five paths should read §25.8. It contains the chapter's central factual correction, and it is short.

💊 GLP-1 — §25.6 is the one to read. The economics of antibiotics are the exact photographic negative of the economics of a chronic metabolic drug, and seeing the contrast explains a great deal about which diseases get new medicines. 🏋️ Performance — §25.3 and §25.5. Membrane-active peptides are sold in the gray market with claims about "natural antibacterial support," and §25.5 is the section that lets you ask the question those sellers cannot answer. 🔬 Science — read straight through. §25.3 and §25.5 are the mechanistic core, and §25.9 is where the book's rating system gets its most careful workout. 💄 Cosmetic — §25.7 matters more than it looks. Topical antimicrobial peptides appear in skincare and wound-care marketing, and the reason topical claims are more defensible than systemic ones is a selectivity argument you can carry directly into Chapter 30. 🏥 Clinical — §25.8 first, then §25.1. Colistin is on your formulary and its story is the most compact illustration of how resistance changes risk-benefit arithmetic that this book contains.

25.1 The resistance crisis, quantified honestly

Before we can evaluate a proposed solution, we need an accurate size for the problem. This is harder than it sounds, and it is a good place to practice a habit that will serve you through the rest of Part V: when a number is doing rhetorical work, find out exactly what it counts.

The best current global estimate comes from the Global Burden of Disease framework. For the year 2019, the analysis estimated that approximately 1.27 million deaths were directly attributable to bacterial antimicrobial resistance, and that resistance was associated with roughly 4.95 million deaths.

Those two numbers are not interchangeable, and the gap between them is where a great deal of sloppy writing lives.

Attributable deaths are the modeled excess: the deaths that would not have occurred if the infecting organism had been susceptible to a first-line drug rather than resistant to it. It is a counterfactual — what would have happened in a world where the resistance was absent but everything else was the same.

Associated deaths are the deaths in which a drug-resistant infection was present in the causal chain at all, including cases where the person would likely have died of the underlying infection even if the organism had been fully susceptible.

Both are legitimate quantities. They answer different questions. The attributable figure asks how much harm does resistance itself cause; the associated figure asks how much death occurs in the presence of resistant infection. This book uses the attributable figure — 1.27 million for 2019 — and says so every time, because the associated figure is the one most often quoted without qualification, and quoting it as though it measured the cost of resistance overstates the case by roughly fourfold.

Notice that this is not a small point of etiquette. If you are going to spend a chapter insisting that a promising therapy be described precisely, you do not get to be loose about the problem it is meant to solve. Precision about the disease is the price of demanding precision about the cure.

🔬 Read the Study

```text FIGURE 25.1 — "How many people does antibiotic resistance kill?" [burden estimate]

THE STUDY A systematic global burden analysis of bacterial antimicrobial resistance for the year 2019, published in 2022. It synthesized hospital records, surveillance systems, literature, and other sources across 204 countries and territories, modeling 23 pathogens and 88 pathogen–drug combinations.

THE QUESTION How much death is caused by antibiotic resistance, globally, in a single year — and how much is merely present alongside it?

WHAT IT SHOWS ~1.27 million deaths directly ATTRIBUTABLE to resistance in 2019. ~4.95 million deaths ASSOCIATED with resistant infection in the same year. Six pathogens — E. coli, S. aureus, K. pneumoniae, S. pneumoniae, A. baumannii, and P. aeruginosa — accounted for the large majority of the attributable total. MRSA alone was linked to more than 100,000 attributable deaths. The burden fell hardest on regions with the least laboratory capacity, which is also where the underlying data are thinnest.

WHAT IT DOESN'T It is a MODELED estimate, not a body count. In many settings there is no routine susceptibility testing at all, so resistance status must be inferred rather than observed. The uncertainty intervals are wide, and they are wider precisely in the places carrying the greatest burden. It is also a snapshot of one year; it does not by itself establish a trend line, and it says nothing about which interventions work.

THE VERDICT The best available global figure, and good enough to establish that the problem is very large. Not precise enough to support arguments that hinge on small differences between years or countries.

THE LESSON When you see a resistance death toll quoted, ask which number it is. If the source does not say "attributable" or "associated," the source does not know, and you should assume the larger and more dramatic figure has been used. ```

Two framings around this number are commonly abused. The first is that the antibiotic pipeline is "empty" — an overstatement, since compounds are in development, but the shape of the concern is right: most new antibacterial agents approved in recent decades have been modifications of existing chemical classes rather than novel mechanisms, and large pharmaceutical companies have largely exited antibacterial discovery since the 1990s. §25.6 explains why, and the explanation is not scientific difficulty alone. The second is the "post-antibiotic era" in which routine surgery becomes impossible. Treat that as a plausible trajectory, not a description of the present. Resistance already makes some infections untreatable today, in specific organisms and specific settings, and the trend is in the wrong direction — which is bad enough without extrapolating to a collapse that has not occurred.

25.2 Antimicrobial peptides as an ancient immune system

Here is a fact that reframes the entire field once you absorb it: essentially every multicellular organism makes antimicrobial peptides. Plants make them. Insects make them. Fish, amphibians, reptiles, birds, and mammals make them. You are making them right now, in your skin, in your airway lining, in your gut, and inside the granules of your neutrophils.

They are part of innate immunity — the fast, non-specific, genetically hardwired arm of host defense that acts within minutes and does not require prior exposure. This is distinct from adaptive immunity, the antibody-and-T-cell system that is slow, specific, and remembers. Adaptive immunity is evolutionarily recent and restricted to vertebrates. Innate immunity, and antimicrobial peptides within it, is far older and far more widely distributed.

The implication is worth stating slowly, because it is the source of the field's optimism and, handled carelessly, the source of its worst overclaim. These molecules have been in continuous use as antibacterial agents for hundreds of millions of years. Whatever they do, they have been doing it against evolving bacteria across geological time, and they still work well enough that natural selection has kept them in essentially every lineage that has tried them.

That is a genuinely remarkable track record and it is the strongest single argument for pursuing this class. It is also — and we will return to this in §25.3 — not the same as the claim that bacteria cannot adapt to them.

What they look like

AMPs are structurally diverse, but three shared features recur so often that they function as a design signature.

Short — typically 12 to 50 residues, squarely inside the peptide size range of Chapter 1, with all the consequences that follow: chemically synthesizable, rapidly cleared, digested if swallowed. Cationic — most carry a net positive charge at physiological pH, from an abundance of the basic side chains lysine and arginine, so the molecule behaves like a small positively charged object in solution. Amphipathic — the crucial architectural feature: the peptide folds, often only upon contact with a membrane, so that hydrophobic residues cluster on one face and charged and polar residues on the other, giving a molecule with a greasy side and a water-loving side.

Put those together and you have described something drawn electrostatically toward a negatively charged surface that can then bury half of itself in a lipid bilayer while the other half stays comfortably in water. That is not a coincidence of chemistry. That is the mechanism, and §25.3 unpacks it.

THE AMPHIPATHIC HELIX — one molecule, two faces

  Viewed down the axis of a helix (a "helical wheel"), residues placed by
  position around the turn. Roughly 3.6 residues per turn (Chapter 1, §1.4)
  means residues spaced 3 or 4 apart in SEQUENCE end up adjacent in SPACE.

                    HYDROPHOBIC FACE
                    (buries in the lipid)
                  Leu   Phe   Ile   Val
                     \   |    |   /
                      \  |    |  /
                   ─────  helix axis  ─────
                      /  |    |  \
                     /   |    |   \
                  Lys   Arg   Lys   Ser
                    CATIONIC / POLAR FACE
                   (stays in water; grips the
                    negatively charged surface)

  A sequence that looks random in one-letter code can be strikingly
  segregated in three dimensions. This is why you cannot evaluate an AMP
  by reading its sequence left to right — you have to know where the
  residues land once the chain coils.

The diagram makes a point that repays attention. The amphipathic arrangement is a property of the folded molecule, and many AMPs are disordered in solution and only adopt a defined structure when they encounter a membrane. This is the "shape acquired on arrival" phenomenon from Chapter 1, §1.4, and it is one reason predicting AMP activity computationally has proven so much harder than it looks. The molecule you model in water is not the molecule that does the killing.

🧬 The Molecule — what "cationic amphipathic" is actually doing

Strip the vocabulary away and an antimicrobial peptide is a very simple device: a positive charge stuck to a grease patch.

The positive charge is a targeting system. Bacterial surfaces are, on the whole, more negatively charged than the outward-facing surface of your own cells. A cationic peptide diffusing through tissue is therefore statistically more likely to end up stuck to a bacterium than to a host cell, not because it recognizes anything about bacteria specifically, but because opposite charges attract. There is no receptor here. There is no binding site, no lock and key, no signal transduction. This is the crudest possible form of molecular targeting, and its crudeness is exactly what makes it interesting.

The grease patch is the weapon. Once the peptide is held against the membrane surface, its hydrophobic face is adjacent to the fatty interior of the bilayer — the most hydrophobic environment available for hundreds of nanometers in any direction. Thermodynamics does the rest. The peptide inserts, and at sufficient local concentration the membrane's integrity fails.

Now hold this next to Chapter 2. Everything you learned there was about receptors: a peptide hormone finds one specific protein, binds it with exquisite selectivity, and triggers an amplified cascade in which one molecule produces thousands of downstream events. Antimicrobial peptides work by almost the exact opposite logic. No receptor, no amplification, no selectivity in the lock-and-key sense — instead, brute physical damage requiring many peptide molecules per target cell.

Two consequences follow immediately, and both matter for everything after this. First, AMPs need to be present at far higher concentrations than hormonal peptides — micromolar rather than picomolar. A hormone whispers; an AMP has to shove. Second, because the targeting is electrostatic rather than structural, the discrimination between a bacterium and one of your own cells is quantitative rather than absolute. Hold onto that sentence. It is §25.5, and §25.5 is the whole problem.

They do more than kill

One complication worth introducing now, because it recurs in Chapter 18 and gets misused constantly: many AMPs have activities beyond direct microbial killing. They recruit immune cells, modulate inflammatory signaling, promote wound closure, and neutralize bacterial products such as lipopolysaccharide. For this reason a growing part of the literature prefers the term host defense peptides, which does not presuppose that killing is the main event.

This is scientifically important and rhetorically dangerous. Important, because some AMPs may prove more useful as immune modulators than as antibiotics, and because their tissue concentrations are sometimes below what direct killing requires in a test tube — which suggests the in vivo job may differ from the in vitro assay. Dangerous, because "it also modulates immunity" is a phrase routinely deployed to rescue a compound whose antibacterial data disappointed. Watch for the pivot: a molecule that failed as an antibiotic and is now described as an immunomodulator may have found its true calling, or may be having its claim quietly moved to ground where the evidence bar is vaguer.

25.3 Membrane disruption, and why resistance is harder — not impossible

Now the mechanism, and the argument that has driven forty years of investment.

Most conventional antibiotics hit a specific molecular target. Beta-lactams inhibit the enzymes that cross-link the cell wall; macrolides, aminoglycosides, and tetracyclines bind specific ribosomal sites; fluoroquinolones inhibit DNA gyrase; rifampicin binds RNA polymerase. In each case activity depends on close, precise contact with one protein or nucleic acid site.

That precision is a strength — it is why these drugs are selective — and a structural vulnerability. A single point mutation in the target can abolish binding. Change one amino acid where the antibiotic docks and the drug's grip fails while the ribosome keeps working. The bacterium survives, divides, and passes the mutation along. Bacterial populations are enormous, generation times short, mutation continuous. Resistance by target modification is the commonest route, and it can arise in a single step.

Most antimicrobial peptides do not have a target in that sense. They have a surface. The sequence of events, for the canonical membrane-disrupting AMP:

1. Electrostatic attraction. The cationic peptide is drawn to the anionic bacterial surface — in Gram-negative organisms, the phosphate groups of lipopolysaccharide in the outer membrane; in Gram-positive organisms, the teichoic and lipoteichoic acids threading the thick peptidoglycan layer.
2. Accumulation and reorientation. Peptide molecules build up on the surface and adopt their amphipathic conformation, hydrophobic face toward the lipid.
3. Insertion. Above a threshold local density, peptides insert into the bilayer.
4. Permeabilization. The membrane loses integrity. Ion gradients collapse, the proton motive force fails, cytoplasmic contents leak, and the cell dies — often within minutes, and importantly, whether or not it is actively dividing.

Three structural models are used to describe step 4, and the field's honest position is that different peptides use different mechanisms and some use more than one.

THREE MODELS OF MEMBRANE DISRUPTION

  BARREL-STAVE          Peptides insert perpendicular to the surface and assemble
                        into a defined channel, hydrophobic faces outward against
   ▓ ▓ ▓ ▓ ▓            the lipid, hydrophilic faces lining a water-filled pore.
   ▓█▓█▓█▓█▓            A discrete structure with a determinate size.
   ▓ ▓ ▓ ▓ ▓            Relatively rare. Gramicidin is the classic example.

  TOROIDAL PORE         Peptides insert but the lipid headgroups bend inward with
                        them, so the pore is lined by BOTH peptide and lipid. The
   ▓╭─╮▓╭─╮▓            membrane curves through itself. More transient, less
   ▓│ │▓│ │▓            regular, and the model best supported for magainins and
   ▓╰─╯▓╰─╯▓            several other well-studied helical AMPs.

  CARPET                Peptides lie flat and accumulate across the surface until,
                        above a critical concentration, the membrane is destabilized
   ▓▓▓▓▓▓▓▓▓▓           wholesale and breaks up in a detergent-like fashion. No
   ══════════           discrete pore at all — the bilayer simply comes apart.
   ▓▓▓▓▓▓▓▓▓▓

  All three share one property that matters more than the differences between them:
  the thing being attacked is a BULK PHYSICAL STRUCTURE, not a molecule with a
  binding site.

That last line is the argument. You cannot mutate your way out of physics as easily as you can mutate your way out of a binding site. A point mutation changes one protein. Resisting a membrane-disrupting agent requires changing the composition, charge, or architecture of the membrane itself — a structure built by many genes, serving many essential functions, and subject to hard biophysical constraints. Restructuring an entire membrane is a much larger evolutionary ask than altering one amino acid in one enzyme.

Add the co-evolutionary argument on top: these peptides have been deployed against bacteria for hundreds of millions of years, and bacteria have not rendered them useless. If resistance to AMPs were as easily acquired as resistance to a beta-lactam, innate immunity would have collapsed as a defense strategy long ago, and it has not.

And now the correction

Everything in the preceding two paragraphs is true, and the way it is usually reported is false.

Bacteria absolutely do evolve resistance to antimicrobial peptides. The mechanisms are known, named, characterized, and in some cases mobile between organisms. A partial list:

Surface charge modification. The most direct counter to an electrostatically targeted weapon is to reduce the charge it aims at. Staphylococcus aureus attaches lysine to a membrane phospholipid, reducing net negative charge and repelling cationic peptides. Many Gram-positives decorate their teichoic acids with D-alanine to the same end. Gram-negatives modify the lipid A of their lipopolysaccharide with aminoarabinose or phosphoethanolamine — and in Salmonella and others this sits under a regulatory system that senses the peptides and switches the modification on.

Efflux. Some organisms pump AMPs back out. Neisseria gonorrhoeae exports the human cathelicidin LL-37 through an efflux system, and efflux-mediated peptide resistance is documented in other genera.

Proteolytic destruction. AMPs are peptides, and bacteria secrete proteases. Several pathogens cleave host defense peptides directly; S. aureus additionally secretes a protein that binds and sequesters LL-37 without cleaving it.

Shielding. Capsules, biofilm matrix, and secreted polysaccharides intercept cationic peptides before they reach the membrane. And, though not strictly bacterial resistance, sequestration by host material is functionally equivalent: many AMPs lose activity in serum, in divalent cations, or at physiological salt — a large part of why in vitro potency so often fails to predict in vivo effect.

And then there is the clinical proof, which is not subtle.

🔬 Read the Study

```text FIGURE 25.2 — "Transferable resistance to a peptide antibiotic" [resistance surveillance]

THE STUDY A routine surveillance program in China screening bacterial isolates from food animals, retail meat, and hospitalized patients for colistin resistance. Reported in 2015. Colistin is a cationic lipopeptide (§25.8) and was at that point one of the last agents reliably active against some multidrug-resistant Gram-negatives.

THE QUESTION Colistin resistance had been seen before, but it was chromosomal — arising by mutation within a lineage and not readily shared. Was there a transferable form?

WHAT IT SHOWS Yes. A plasmid-borne gene, designated mcr-1, encoding an enzyme that adds phosphoethanolamine to lipid A. The modification reduces the negative charge of the outer membrane, weakening the electrostatic attraction that colistin depends on. Because the gene sits on a plasmid, it can move between bacteria — including between species — rather than only being inherited down a single lineage. Within roughly a year of the report, mcr-1 had been identified in isolates from many countries across several continents.

WHAT IT DOESN'T It does not show that colistin has stopped working; colistin remains clinically useful. It does not quantify how much mcr-mediated resistance contributes to patient outcomes. And it concerns one specific cationic lipopeptide — it is not a direct demonstration that resistance to every membrane-active peptide is equally transferable.

THE VERDICT Definitive on the narrow point, and the narrow point is important: resistance to a membrane-targeting cationic peptide can be encoded on a mobile genetic element and spread horizontally between organisms.

THE LESSON This is the single fact that should be attached to every claim that AMPs are resistance-proof. The mechanism-based argument for AMPs — that membranes are hard to redesign — predicts that resistance will be SLOWER and COSTLIER to evolve. It does not predict that it will not happen, and here it demonstrably has. ```

So what is the accurate claim? It is this: resistance to membrane-active antimicrobial peptides appears to be harder to evolve, and to carry higher fitness costs, than resistance to target-specific antibiotics. That is a meaningful advantage and it is worth pursuing. It is not immunity, it is not permanence, and it is not a guarantee.

This is the chapter's main hype risk, and it is worth naming explicitly. The claim "AMPs don't generate resistance" is attractive because it converts a quantitative advantage into a qualitative one, and qualitative claims make better headlines. It also, conveniently, promises to end the problem described in §25.1 rather than merely to buy time against it. Be suspicious of any account of this field that reaches for the stronger version.

📊 Evidence Rating

Claim: Antimicrobial peptides cannot generate bacterial resistance, because membrane disruption is not a mutable target. Rating: ❌ Hype outpaces evidence (assessed as of this writing, 2026) Reason: Multiple well-characterized resistance mechanisms exist — surface charge modification, efflux, proteolysis, shielding — and at least one, plasmid-borne mcr, is transferable between organisms and has spread internationally. The defensible claim is that resistance is harder and costlier to evolve, not that it is impossible. What would change it: Nothing available could rescue the claim as stated; it is contradicted by direct observation. The weaker claim — that resistance to membrane-active peptides emerges more slowly and imposes larger fitness costs than resistance to target-specific agents — is supported in some systems and would be strengthened by long-term clinical use of a systemic AMP with resistance surveillance, which does not yet exist to study.

Note the shape of that rating, because §25.9 will formalize it. The ❌ is not aimed at antimicrobial peptides. It is aimed at one sentence about antimicrobial peptides. The molecules themselves are about to receive a 🔬 in the same chapter, and there is no contradiction: one molecule, many ratings is the fifth rule of the system introduced in Chapter 5.

25.4 Defensins, cathelicidins, magainins

Three families do most of the work in the human-facing literature. Knowing them is enough to read almost any AMP paper without getting lost.

Defensins

Defensins are small cationic peptides, characteristically stabilized by three disulfide bonds (Chapter 1, §1.2 — cysteine's sulfur staple), and folded into a compact beta-sheet structure. Humans make two classes.

α-defensins are found in the granules of neutrophils, where they are among the most abundant proteins present, and in the Paneth cells at the base of the intestinal crypts. The neutrophil versions are released into the phagosome when the cell engulfs a bacterium — a strikingly sensible piece of engineering, since it puts an indiscriminate membrane-disrupting agent inside a sealed compartment where it can do its work without touching the host cell around it. Hold that thought; it is a preview of §25.5, and it is essentially nature's answer to the selectivity problem.

β-defensins are produced by epithelial cells — skin, airway, gut, urogenital tract. Some are made constitutively; others are induced by infection or inflammation. They are part of the reason an intact epithelial surface is a functioning antimicrobial barrier rather than merely a physical one.

Defensins also have signaling roles, including chemotactic activity for immune cells. They are, in other words, host defense peptides in the fuller sense described at the end of §25.2.

Cathelicidins

Cathelicidins are defined by a conserved precursor region — the cathelin domain — rather than by the structure of the active peptide, which varies widely between species. The precursor is stored inactive and cleaved to release the mature antimicrobial peptide when needed, which is a recurring theme in peptide biology and one you met in Chapter 2: store the safety catch, release on demand.

Humans have exactly one. It matters enough to get its own box.

🧬 The Molecule — LL-37, the only human cathelicidin

Identity. A 37-residue cationic, amphipathic, largely α-helical peptide. Its name is a description: it begins with two leucine residues and is 37 amino acids long. It is released by proteolytic cleavage from a stored precursor protein, and it is produced by neutrophils, epithelial cells throughout the body, and several other cell types.

What it does — the part everybody agrees on. It permeabilizes bacterial membranes in the manner described in §25.3, with activity against both Gram-positive and Gram-negative organisms in vitro. It also binds and neutralizes lipopolysaccharide, blunting the inflammatory response to Gram-negative bacterial debris.

What it does — the part that complicates the story. LL-37 is chemotactic for neutrophils, monocytes, and T cells; influences epithelial cell migration and wound repair; affects angiogenesis; and modulates responses to microbial products through multiple pathways. Its expression is upregulated by vitamin D signaling, which is the mechanistic thread behind a large literature on vitamin D and infection — a literature whose clinical results have been considerably less impressive than the mechanism suggested, and which Chapter 18 handles in more detail.

Why it is a cautionary tale as much as a hero. LL-37 has been implicated on both sides of the ledger. It contributes to antibacterial defense at epithelial surfaces. It has also been implicated in inflammatory disease processes, including in psoriasis, where complexes of LL-37 with host DNA are thought to drive an inappropriate immune activation. The same molecule, in the same body, is protective in one context and pathogenic in another — which is a general property of innate immune effectors and a specific warning against any product marketed on the premise that boosting a host defense peptide is straightforwardly good.

The rating discipline applies here too. "LL-37 kills bacteria in culture" is well established. "LL-37 supplementation improves human infection outcomes" is a completely different claim, with a completely different evidence base, and belongs to a different rating. Chapter 5's first rule — a rating attaches to a claim, with a population and an endpoint, never to a molecule — does not stop applying because the molecule is one your own body makes.

Magainins

Magainins are α-helical peptides isolated from the skin of the African clawed frog, Xenopus laevis, and described in 1987. The observation that prompted the search was the one in this chapter's opening: frogs recovering from surgery in non-sterile water did not develop wound infections.

Their historical importance exceeds their current clinical importance, and that is exactly why they belong here. Magainins were the compounds that convinced the pharmaceutical world that antimicrobial peptides were drug candidates rather than immunological curiosities. They are structurally simple, they kill a broad range of organisms in vitro, they were tractable to synthesize and to modify, and they became the workhorse system for studying how membrane-active peptides do what they do — much of the toroidal pore model comes from magainin work.

A synthetic magainin analog, pexiganan, went further into clinical development than almost any other AMP. What happened to it is §25.7, and it is the most instructive failure in the field.

Three families, three lessons. Defensins show the containment strategy — kill inside a sealed phagosome. LL-37 shows that AMPs are not only antibiotics, for better and worse. Magainins show that the field's founding molecules were beautiful in vitro and did not become drugs, which is the question the rest of the chapter answers.

25.5 The selectivity problem: our membranes are membranes too

This is the section the chapter is built around. If you remember one thing from Chapter 25, make it this one, because it explains, in a single mechanism, why forty years of excellent science has not produced a systemic drug.

An antimicrobial peptide kills by disrupting a lipid bilayer. You are made of lipid bilayers.

The field's answer to this — and it is a real answer, not a hand-wave — is that bacterial and mammalian membranes differ. They differ in three ways that matter:

Charge. Bacterial cytoplasmic membranes are rich in anionic phospholipids — phosphatidylglycerol, cardiolipin — and their outer surfaces carry additional negative charge from lipopolysaccharide (Gram-negative) or teichoic acids (Gram-positive). Mammalian plasma membranes present mostly zwitterionic phospholipids on their outer leaflet — phosphatidylcholine and sphingomyelin — which carry both a positive and a negative charge and are therefore net neutral. The anionic phospholipids in your cells are largely sequestered on the inner leaflet, facing the cytoplasm, where a peptide approaching from outside never sees them.

Cholesterol. Mammalian membranes contain cholesterol, which packs between phospholipid tails and increases the ordering and mechanical rigidity of the bilayer. This makes insertion by an amphipathic peptide thermodynamically less favorable. Bacterial membranes generally lack cholesterol.

Transmembrane potential. Bacterial membranes typically maintain a larger potential difference, inside negative, which helps draw cationic peptides in.

WHY THE PEPTIDE PREFERS THE BACTERIUM — and why "prefers" is the operative word

  BACTERIAL OUTER SURFACE              MAMMALIAN OUTER SURFACE
  ───────────────────────              ───────────────────────
   ⊖  ⊖  ⊖  ⊖  ⊖  ⊖  ⊖                  ⊕⊖  ⊕⊖  ⊕⊖  ⊕⊖  ⊕⊖
   anionic phospholipids                zwitterionic phospholipids
   + LPS or teichoic acids              (net neutral overall)
   no cholesterol                       cholesterol between the tails
   large membrane potential             smaller potential
   ─────────────────────────            ─────────────────────────
   STRONG electrostatic pull            WEAK electrostatic pull
   LOOSE, insertable bilayer            ORDERED, rigid bilayer

                     ⟹  the peptide accumulates faster,
                        and inserts more readily, on the left

  BUT: this is a difference in DEGREE. There is no molecular feature on the
  bacterial side that is absent on the mammalian side in a binary way. Raise the
  concentration high enough and the peptide goes into your membranes too.

The distinction is a matter of degree, not of kind. That sentence is the whole obstacle. A receptor-mediated drug can be selective in a binary sense — the receptor is present on one cell type and absent on another, and the drug does nothing where it is missing. A membrane-active peptide has no such off switch. It has a preference, and preferences can be overwhelmed by concentration.

Hemolysis and the selectivity index

Three terms appear in every AMP paper. Minimum inhibitory concentration (MIC) is the lowest concentration preventing visible bacterial growth under defined conditions — the standard measure of antibacterial potency, lower being better. Hemolysis is the lysis of red blood cells, measured by incubating the peptide with erythrocytes and quantifying released hemoglobin; red cells are used because they are easy to obtain, easy to read out, and — being membrane sacs without much else going on — a clean test of pure membrane damage. The selectivity index is the ratio of the concentration that harms host cells to the concentration that kills bacteria. Higher is better. It is this field's version of the therapeutic window, and it is the number on which candidates live and die.

THE SELECTIVITY INDEX — the arithmetic that decides everything
                        (numbers below are INVENTED for illustration;
                         they are not measurements from any study)

  PEPTIDE A          kills bacteria at        4 units
                     lyses red cells at     256 units
                     selectivity index = 256 / 4  =  64      ← wide margin

  PEPTIDE B          kills bacteria at        4 units
                     lyses red cells at      16 units
                     selectivity index =  16 / 4  =   4      ← no room to dose

  Both peptides are EQUALLY POTENT against bacteria. Their MICs are identical.
  A paper reporting only MIC would present them as equivalent discoveries.
  Only one of them could conceivably become a systemic drug, and the number
  that separates them is not in the antibacterial data at all.

That illustration is the practical heart of the section. A great many otherwise excellent antimicrobial peptides have failed on this single axis. They kill bacteria beautifully. They also damage host cells at concentrations uncomfortably close to their antibacterial concentrations, and when you account for what happens in real tissue — binding to serum proteins, inhibition by physiological salt and divalent cations, degradation by host and bacterial proteases — the concentration you need in the body rises while the concentration your body tolerates does not.

The margin closes from both directions. This is the single best explanation for the field's central puzzle, which is that thousands of peptides show excellent in vitro activity and almost none have become systemic drugs.

🩺 Safety and Risk — what "selectivity failure" looks like in a person

The selectivity index is an in vitro number and it is a screening tool, not a prediction. Peptides with acceptable-looking indices have still failed in humans, and the failures are informative.

The kidney is the recurring casualty. Peptides are cleared substantially by the kidney, and the proximal tubule cells that reabsorb filtered material end up concentrating them. A peptide that is comfortably tolerated by cells in a dish can accumulate in exactly the tissue least able to escape it. Murepavadin, a peptide-derived antibacterial targeting Pseudomonas aeruginosa, reached Phase 3 trials for intravenous use and had those trials halted in 2019 after signals of acute kidney injury. This was not a failure of antibacterial activity — the compound worked against its target organism. It was a failure of the margin between working and harming, discovered late and at considerable cost.

Nephrotoxicity is also why the peptide antibiotics we already have are used the way they are. §25.8 covers colistin, whose kidney toxicity is well documented and whose place in therapy is constrained by it. That drug is in use not because its safety profile is good but because the alternative in some cases is an infection with no treatment at all.

The gray-market angle. Compounds described as "antimicrobial peptides" occasionally appear in unregulated channels with claims about natural infection defense. Apply §25.5 directly: the mechanism is indiscriminate membrane disruption, the margin between antibacterial and cytotoxic concentrations is narrow enough to have defeated well-resourced clinical programs, and no seller of an unapproved preparation has characterized that margin for the material in the vial. Even a correctly identified, pure antimicrobial peptide would carry an unresolved toxicity question.

As always, decisions about treating an infection belong with a clinician. An untreated bacterial infection is a medical emergency in a way that most topics in this book are not.

Why nature does not have this problem — and why we do

Your body deploys these molecules constantly without lysing its own cells, and every trick it uses is one a systemic drug cannot copy. Compartmentalization: neutrophil defensins are released into the phagosome, a sealed vesicle containing the bacterium and essentially nothing of yours, where the peptide reaches lethal concentration in a few cubic micrometers and never meets your membranes at all. Locality: epithelial AMPs act on the outward face of a barrier, in mucus, and the tissue behind never sees high concentrations. Regulation: expression is induced where and when needed, then stops. Cooperation: AMPs work alongside complement, phagocytes, antibodies, and mucociliary clearance, and are never asked to sterilize a bloodstream single-handedly.

An intravenous drug has none of these advantages. It goes everywhere, at once, at whatever concentration is needed at the most inaccessible site of infection, for as long as therapy lasts. You are asking a molecule evolved for sealed compartments and epithelial surfaces to work as a systemic agent, and the thing that made it safe in its native context — that it never got far from where it was released — is precisely what you have taken away.

🔍 Check Your Understanding

1. Two AMPs have identical MICs against E. coli. One has a selectivity index of 60 and the other of 3. What does that tell you, and which piece of information would a headline about "a powerful new bacteria-killing peptide" most likely omit?
2. Why is a neutrophil able to use an indiscriminate membrane-disrupting peptide safely, and which of its advantages is available to an intravenous drug?
3. Explain, without using the word "selectivity," why topical use of an AMP faces a lower bar than intravenous use.

25.6 Manufacturing cost and the economics of antibiotics

Suppose the selectivity problem were solved tomorrow. Suppose a peptide emerged with a wide margin, a clean kidney profile, and broad activity against resistant Gram-negatives.

It would still be very difficult to bring to market, and the reasons have nothing to do with biology. This section is about the second obstacle, which is at least as large as the first and is the one most often left out of enthusiastic coverage.

Cost of goods

Peptides are expensive to make. Chapter 32 covers manufacturing in detail; the short version is that solid-phase synthesis builds a chain one residue at a time, each step requiring reagents, solvent, and purification, with yield losses compounding across the sequence — so a 30-residue peptide costs considerably more than three times a 10-residue one. Recombinant production in engineered cells is an alternative for some sequences, but antimicrobial peptides present an obvious problem for that route: they kill the organism you are trying to produce them in, requiring expression as an inactive fusion partner plus a cleavage step, which adds cost of its own.

For a drug taken daily for the rest of a patient's life, a high cost of goods is absorbable. For an antibiotic it is not, and the reason is the next subsection.

The market failure

Antibiotics are, commercially, the worst drug class in medicine. This is not a complaint about greed; it is arithmetic, and the arithmetic is counterintuitive enough to walk through.

A commercially successful drug is taken by many people, for a long time, at a sustainable price. A chronic metabolic drug fits perfectly — a patient may take it daily for decades. Now consider a new antibiotic, and specifically a good one: a genuinely novel agent active against organisms nothing else touches.

It is used briefly. A course of antibiotics lasts days, not years.

It is used rarely. Most infections are treated with older, cheaper, well-characterized drugs.

It is deliberately withheld. Here is the part that makes antibiotics unique. The better the new drug, the harder infectious disease specialists will work to prevent its use. A novel agent against resistant organisms is placed under stewardship restriction and reserved for cases where nothing else will do — because using it widely will select for resistance and destroy its value for everyone. This is correct medicine. It is also, from the manufacturer's point of view, a formal program to suppress sales of the product.

And the reward for success is obsolescence. If the drug is used, resistance eventually emerges, and the asset depreciates in a way no other drug class experiences.

TWO REVENUE CURVES

  CHRONIC THERAPY (e.g., a metabolic drug)
  revenue │                        ▁▃▅▇█████████████
          │                  ▁▃▅▇██
          │            ▁▃▅▇██
          │      ▁▃▅▇██
          └──────────────────────────────────────────── time
            launch → uptake → years of daily use per patient

  NOVEL ANTIBIOTIC (reserved, stewarded)
  revenue │
          │      ▁▂▁▂▁▁▂▁▂▁▁▂▁▂▁▁▂▁▂▁▁▂▁▂▁▁▂▁▂▁▁▂▁▂▁▁
          │    ▁▂
          └──────────────────────────────────────────── time
            launch → restricted formulary → sparing use → patent expiry

  The second curve is the intended outcome of good stewardship. The clinical
  system is functioning exactly as designed, and the commercial result is a
  product that cannot repay what it cost to develop.

This is not theoretical. A company that won approval for a novel antibacterial agent against multidrug-resistant Gram-negative infections in 2018 filed for bankruptcy less than a year later, having been unable to generate meaningful revenue from a drug that did what it said. Another antibiotic-focused company entered bankruptcy in the same period. Approval, in this class, is not the finish line — and several firms have discovered that after crossing it.

The consequence is that most large pharmaceutical companies have exited antibacterial discovery, and much of what remains is done by small firms, academic groups, and public-private partnerships that depend on grant funding to survive the years before approval and, increasingly, after it.

What has been proposed

Because the problem is structural, the proposed solutions are structural. They share a common insight: if the problem is that revenue is tied to volume, break the link between revenue and volume.

Subscription-style procurement, sometimes described by analogy to a streaming service: a health system pays a fixed annual sum for access regardless of how much is used, so the manufacturer is paid for availability rather than units. Pilot programs along these lines have been implemented in at least one national health system.

Market entry rewards. A large lump-sum payment on approval of a qualifying novel antibiotic, decoupled from sales entirely. A prominent international review of antimicrobial resistance published in 2016 recommended rewards on the order of a billion dollars per qualifying drug.

Extended or transferable exclusivity, granting the developer additional market exclusivity or a voucher applicable to another product — repeatedly proposed and repeatedly criticized, since the cost falls on payers of an unrelated medicine. And push funding, direct subsidy of early research, which already exists through public and philanthropic mechanisms and is why much of the current pipeline exists at all.

This book takes no position on which of these is right. They involve trade-offs in public spending, intellectual property, and international coordination that are outside the scope of a book about peptides, and reasonable people disagree. What is not in dispute is the diagnosis: the antibiotic market does not reward the drugs that are most needed, and no scientific advance changes that. A reader who finishes this chapter believing that antimicrobial peptides will arrive as soon as the biology is worked out has missed half the problem.

⚠️ Hype Check — "the industry is suppressing antimicrobial peptides because you can't patent nature"

The claim, in its usual form:

"AMPs are natural molecules your body already makes, so no company can patent them. That's the real reason we don't have peptide antibiotics — there's no money in a cure."

What's true in it. The economics genuinely are broken, and that genuinely is why large companies have left the field. If someone arrives at "the market is failing" from this claim, they have reached a correct conclusion, and this section agrees with them.

Where it fails. In almost every particular.

First, the patent premise is wrong. Essentially no AMP entering development is the natural sequence; candidates are engineered analogs with substituted residues, altered charge, D-amino acids, cyclization, or lipid attachment — and modified sequences, formulations, processes, and uses are all patentable. Pexiganan, murepavadin, and omiganan were proprietary compounds owned by companies that expected to profit from them.

Second, the suppression premise fails on the evidence of failure itself. Companies spent large sums taking these compounds into Phase 3. You do not run a Phase 3 trial of a molecule you are suppressing. The programs stopped because the trials did not show what was needed, or the toxicity did.

Third, and most importantly, the claim erases §25.5. There is a specific, well-characterized, physical reason systemic AMPs are hard, and attributing the difficulty to a conspiracy requires ignoring a large published literature that describes that difficulty in detail. Fourth, the claim is refuted by §25.8: peptide antibiotics are on the market. Somebody is selling daptomycin.

Verdict: the conclusion is half right for entirely wrong reasons. There is a market failure, it is severe, and it is well documented — and it operates on top of a genuine scientific obstacle, not instead of one. Claims that replace a hard technical problem with a villain should always prompt you to go looking for the technical problem, because it is usually still there.

25.7 What is in trials, and the topical approvals that exist

Now the status report, delivered plainly and with a date attached.

As of this writing, in 2026, no antimicrobial peptide has become a widely used systemic antibiotic. Several have entered clinical trials. Several have reached Phase 3. None has emerged as a broad-spectrum injectable agent in general use. That statement will need revisiting; the mechanism by which you should revisit it is §25.9.

Topical applications have gone considerably further than systemic ones, and this follows directly from §25.5. A topical agent faces a much lower selectivity bar for three reasons. The exposed tissue is limited — a peptide applied to intact or wounded skin reaches a small, defined volume rather than every cell in the body. Systemic absorption is low, since a peptide is large, charged, and poorly absorbed across skin, which is a delivery obstacle in Chapter 30's context and a safety advantage here. And the consequences of local irritation are tolerable: redness or stinging at an application site is an adverse effect to weigh against benefit, while acute kidney injury is a different category of problem.

So the pattern in the clinical record is not an accident of history. It is the selectivity index expressing itself in the structure of the development pipeline.

That said, "further" is not "successfully," and the most instructive story in the field is a topical one.

🔬 Read the Study

```text FIGURE 25.3 — "The most advanced antimicrobial peptide trial, twice" [Phase 3, failed]

THE STUDY Pexiganan, a synthetic analog of magainin (§25.4), formulated as a topical cream for mild infection of diabetic foot ulcers. Developed over roughly two decades. An FDA advisory committee reviewed it in 1999 and it was not approved. It was later revived by another company, which ran two Phase 3 trials reporting in 2016.

THE QUESTION Does topical pexiganan improve clinical outcomes in mildly infected diabetic foot ulcers, compared with the relevant comparator?

WHAT IT SHOWS In the 1999 review, the central problem was not that the compound did nothing — it was that a benefit over standard oral antibiotic treatment was not demonstrated to the agency's satisfaction. In the 2016 trials, the compound again failed to demonstrate the required benefit; response rates were high in the comparator arm as well, leaving no separation to detect.

WHAT IT DOESN'T It does not show that pexiganan lacks antibacterial activity — it has it, abundantly, in vitro. It does not show that AMPs cannot work topically. And a trial that fails because BOTH arms did well is a specific kind of failure: it tells you the drug did not beat the comparator in that population, not that it did nothing.

THE VERDICT Negative for the claim tested. The most advanced AMP program of its era did not produce an approved product across two separate development efforts and roughly twenty years.

THE LESSON Two, and both are general. First: in vitro potency is not efficacy. Pexiganan's killing data were never the problem. Second: a trial is a comparison, and the comparator matters as much as the drug. An agent that would look impressive against no treatment can fail completely against good standard care — and choosing an easy comparator is one of the commonest ways a weak drug is made to look strong in the literature you will read in Part VI. ```

Other programs worth knowing about, in outline:

Omiganan, an analog of a bovine host defense peptide, was studied topically for catheter-site infection and later for dermatologic conditions, through multiple trials, without producing a broadly used product. Murepavadin (§25.5) is peptide-derived but target-specific rather than purely membrane-disruptive — notable because it shows the field diversifying away from lysis; its intravenous Phase 3 program was halted in 2019 for kidney toxicity, and later work has explored inhaled administration, itself a selectivity strategy. Defensin-mimetic and other peptidomimetic compounds attempt to keep the mechanism while escaping the peptide's liabilities — proteolysis, cost, rapid clearance — an approach Chapter 33 covers in general.

Teixobactin, reported in 2015 and discovered through a technique for culturing previously uncultivable soil bacteria, is a depsipeptide that binds lipid precursors of the bacterial cell wall rather than disrupting membranes directly. It attracted enormous attention because resistance was not readily generated in laboratory experiments. It has not, as of this writing, completed human clinical trials, and it belongs in the 🔬 column until humans have been dosed and the results published.

And here is the part that most coverage omits: antimicrobial peptides are already approved and in use topically, and have been for decades. Bacitracin is a cyclic peptide and a staple of over-the-counter wound ointment. Polymyxin B is a cyclic lipopeptide and appears in the same products, as well as in ophthalmic and otic preparations. Gramicidin is a peptide and is used in some topical eye preparations. Nisin, a peptide produced by a lactic acid bacterium, is used as a food preservative and has been for many years.

The tube of triple antibiotic ointment in a bathroom cabinet is, in substantial part, a peptide product. That is not a rhetorical flourish; it is an accurate description of the composition, and it should make you suspicious of any article that presents topical antimicrobial peptides as an emerging technology.

📊 Evidence Rating

Claim: Topical antimicrobial peptide preparations can prevent or treat localized bacterial infection in specific applications. Rating: ⚠️ Promising but preliminary (assessed as of this writing, 2026) Reason: Several peptide agents are approved and in long-standing topical use, and multiple randomized trials of newer topical AMPs have been conducted — but the most advanced program, pexiganan, failed to demonstrate benefit over comparator in two separate development efforts, and the evidence across the category does not settle which topical indications are genuinely improved by AMP-based products versus adequately served by existing care. What would change it: Adequately powered randomized trials of a specific topical AMP in a specific indication, against an appropriate active comparator, showing a consistent clinical benefit — would move that claim to ✅. Continued failures against active comparators would move it toward ❌ for those specific uses.

Notice how much narrower that rating is than "topical antimicrobial peptides work." It has to be. Bacitracin's decades of use tells you nothing about whether a novel engineered AMP cream will improve diabetic foot ulcer outcomes, and pexiganan's failure tells you nothing about bacitracin. One category, many claims, many ratings.

25.8 Colistin and daptomycin: the peptide antibiotics already in use

If this chapter has an argument that will change how you read the field, it is here.

The category is not speculative. Peptide antibiotics have been in clinical use for over sixty years. They are on hospital formularies, in treatment guidelines, in the bloodstreams of patients today. What has not been achieved is something narrower: a new generation of broad-spectrum systemic antibiotics derived from the host defense peptides of §25.2 and §25.4. Those are different statements, and conflating them produces the two symmetrical errors this book exists to correct — the enthusiast's "a whole new class of antibiotics is coming" and the cynic's "peptide antibiotics have never worked."

Colistin (polymyxin E)

What it is. A cyclic lipopeptide produced by a soil bacterium, discovered in the late 1940s. It is not a ribosomally synthesized peptide of the kind described in §25.2 — it is made by bacterial enzymes and includes non-standard amino acids — but it is unambiguously a peptide, and it is cationic, amphipathic, and membrane-disrupting. It works exactly as §25.3 describes: the positively charged peptide binds the negatively charged lipid A of Gram-negative lipopolysaccharide, displaces the divalent cations that stabilize the outer membrane, permeabilizes it, and reaches the inner membrane, which it also disrupts.

Its history is the point. Colistin entered clinical use around 1960 and was largely abandoned by the 1970s and 1980s, because it caused kidney injury and neurological effects, and because a new generation of aminoglycosides and later beta-lactams offered comparable Gram-negative coverage with better tolerability. It was, for practical purposes, a retired drug.

It came back. Beginning in the 2000s, as carbapenem-resistant Klebsiella, multidrug-resistant Acinetobacter baumannii, and resistant Pseudomonas aeruginosa became more common, clinicians found themselves facing infections against which nothing in the modern armamentarium worked — and reached back for the drug they had shelved. Colistin returned as a last-line agent against multidrug-resistant Gram-negative organisms, and remains one.

💊 In the Clinic — what colistin's return actually demonstrates

A drug is not withdrawn from practice lightly, and it is not reinstated lightly either. Colistin's trajectory is worth sitting with, because it is the clearest available illustration of something that is otherwise abstract: risk-benefit is a ratio, and resistance changes the denominator.

In 1980, a physician weighing colistin against an aminoglycoside for a susceptible Gram-negative infection was comparing a drug with substantial nephrotoxicity and neurotoxicity against a drug with less of both and similar efficacy. The choice was easy, and colistin lost it.

In 2015, a physician facing a carbapenem-resistant organism was comparing colistin's toxicity against an untreated infection with high mortality. Nothing about colistin had changed. The comparison had. The same drug, with the same side-effect profile, moved from unacceptable to essential without a single new molecule of evidence about the drug itself.

Two things follow. First, "is this drug safe?" is never a complete question. Safe compared to what, in whom, with what alternatives? A tolerability profile that is disqualifying for a routine infection is acceptable for an otherwise untreatable one — the reasoning that governs oncology, transplant medicine, and infectious disease, and that is invisible in most consumer discussion of medicine. Second, and less comfortably: colistin's revival is a measure of how bad the resistance problem has become. We are using a drug we had already decided was too toxic, because the alternative became worse. That is not a triumph of pharmacology; it is a report on §25.1.

Colistin is used under specialist supervision with monitoring of kidney function, and it is administered as an inactive prodrug converted to the active peptide in the body. Details of its use are a clinical matter and are deliberately not given here.

📊 Evidence Rating

Claim: Colistin is an effective treatment option for infections caused by multidrug-resistant Gram-negative organisms when other agents are unavailable or inactive. Rating: ✅ Strong clinical evidence (assessed as of this writing, 2026) Reason: Approved, in guideline-directed clinical use for decades, with well-characterized activity against organisms including carbapenem-resistant Enterobacterales, Acinetobacter, and Pseudomonas; its role as a last-line agent is established practice, and its toxicities are documented well enough to be actively monitored for. What would change it: Widespread dissemination of transferable resistance such as mcr to the point of routine clinical failure; or head-to-head trials showing newer agents are superior for the same organisms, which would narrow rather than eliminate the claim.

Read that rating carefully, because it is doing something unusual. Colistin causes kidney injury in a substantial fraction of patients who receive it, and it can cause neurological effects including neuromuscular blockade. Those are not footnotes; they are the reason the drug was abandoned once already. And the rating is still ✅.

This is what ✅ means and what it does not mean. It means the evidence supports the claim as stated — that this drug treats these infections. It does not mean the drug is pleasant, well tolerated, or the first thing anyone would want. A rating is an assessment of evidence, not an endorsement, and a reader who reads ✅ as "good drug" rather than "supported claim" will misread this book from here on. Some of the most firmly established treatments in medicine have side-effect profiles that would be intolerable in any other context, and the reason they are used anyway is the denominator argument in the box above.

Daptomycin

What it is. A cyclic lipopeptide produced by a soil actinomycete, active against Gram-positive organisms including methicillin-resistant Staphylococcus aureus and vancomycin-resistant enterococci. It was approved in 2003 for complicated skin and skin structure infections, and in 2006 for S. aureus bloodstream infection including right-sided endocarditis. It is in routine hospital use.

How it works is a variation on §25.3 with a twist: daptomycin requires calcium, which it binds to acquire the positive charge needed to interact with the anionic Gram-positive membrane. Once inserted, it oligomerizes and disrupts membrane function, causing loss of membrane potential and cell death. The calcium dependence is why its activity in the laboratory depends on the calcium content of the test medium — a detail that has produced real confusion in susceptibility testing.

And it has a beautiful limitation. Daptomycin is inactivated by pulmonary surfactant, the lipid-rich material lining the alveoli, which binds and neutralizes it. Consequently daptomycin is not used for pneumonia — the drug works everywhere else in the body and is defeated by the biochemistry of the lung. This is one of the most elegant illustrations available of a lesson from Chapter 4: a drug's effect depends on what happens to it at the site of action, and a molecule can be potent, well-tolerated, and completely useless in one specific tissue.

📊 Evidence Rating

Claim: Daptomycin is effective for its approved indications — complicated skin and skin structure infections, and S. aureus bloodstream infection including right-sided endocarditis. Rating: ✅ Strong clinical evidence (assessed as of this writing, 2026) Reason: Approved on the basis of randomized controlled trials, in routine clinical use since 2003, with a characterized safety profile and established resistance monitoring. What would change it: Accumulating evidence of clinical failure in these indications, or resistance prevalence high enough to make empiric use unreliable. Note that the rating does not extend to pneumonia, where the drug is inactivated by surfactant — a separate claim with a different and negative answer.

And the rest of the shelf

Vancomycin is a glycopeptide — a peptide backbone bearing sugars — approved in the 1950s and still a mainstay of Gram-positive therapy. Its mechanism is not membrane disruption: it binds a structural component of cell wall precursors and blocks cross-linking, a reminder that "peptide antibiotic" describes composition, not mechanism. Bacitracin, polymyxin B, and gramicidin appear in topical preparations, as noted in §25.7. And teicoplanin, dalbavancin, oritavancin, and telavancin are (lipo)glycopeptides in the vancomycin family, several approved in recent decades — so new peptide antibacterials have in fact reached market within living memory, just not from the host-defense-peptide lineage.

THE PEPTIDE ANTIBIOTICS ALREADY IN USE

  AGENT              CLASS                  ROUTE        STATUS
  ────────────────   ────────────────────   ──────────   ─────────────────────────
  Colistin           cyclic lipopeptide     systemic     last-line, MDR Gram-negatives
  Polymyxin B        cyclic lipopeptide     both         topical + restricted systemic
  Daptomycin         cyclic lipopeptide     systemic     routine, Gram-positive
  Vancomycin         glycopeptide           systemic     mainstay, Gram-positive
  Teicoplanin et al. (lipo)glycopeptides    systemic     approved, in use
  Bacitracin         cyclic peptide         topical      over-the-counter
  Gramicidin         linear peptide         topical      ophthalmic preparations
  Nisin              lantibiotic            n/a          food preservative

  NONE of these came from the host-defense-peptide research program of §25.2–25.4.
  All of them are peptides that kill bacteria, in clinical or commercial use, today.

Finish this section with the corrected sentence. Not: antimicrobial peptides don't work as drugs yet. Instead: peptide antibiotics work, are approved, and are in daily use — and the specific project of converting the host defense peptides of innate immunity into new broad-spectrum systemic antibiotics has not yet produced an approved agent, for reasons that are well understood and are partly biological and partly economic.

The second sentence is longer, less quotable, and true.

25.9 Writing a 🔬 verdict honestly

Chapter 5 introduced four ratings. Most of this book's work happens in the difference between ✅ and ⚠️, or between ⚠️ and ❌. This chapter is where the fourth rating earns its place, and it is the one readers most often misuse.

🔬 and ❌ can look similar from a distance. In both cases, the honest answer to "does this work in humans?" is "we do not know." The difference is in what produced that state.

❌  or  🔬 ?  — the discriminating question

  START: someone is making a confident claim about a human outcome.

    │
    ├─ Is there human evidence?
    │     │
    │     ├─ YES, and it contradicts the claim ───────────────────► ❌
    │     │
    │     ├─ YES, real randomized data, but it does not settle
    │     │   the question ──────────────────────────────────────► ⚠️
    │     │
    │     └─ NO, or only preclinical work
    │           │
    │           ├─ …and the claim is being made CONFIDENTLY
    │           │   anyway, marketed, sold, asserted as known ───► ❌
    │           │
    │           └─ …and nobody is claiming otherwise; the work is
    │               preclinical or early-phase and is DESCRIBED
    │               as such by the people doing it ──────────────► 🔬

  ❌ is a judgment about a CLAIM that has outrun its evidence.
  🔬 is a description of a FIELD that is proceeding properly and is not finished.

  Both mean "we don't know." Only one of them is anybody's fault.

The distinction is not decoration. ❌ means a confident claim is unsupported — usually because human evidence is absent while the claim is being made anyway, or because human evidence exists and went the other way. It is a criticism, and the criticism is directed at the assertion, not at the molecule.

🔬 means the science is early, proceeding properly, and it is too soon to rate. No one has overclaimed. The work is being done in the right order. The obstacles have been identified and described by the researchers themselves. There is simply no clinical answer yet, and the honest posture is to say so.

Antimicrobial peptides as systemic antibiotics are this book's cleanest 🔬, and it is worth being explicit about why each element is satisfied:

A real mechanism. Not a plausible story — a mechanism demonstrated at the biophysical level, observable in real time, understood well enough to be modeled in three competing structural frameworks that make different testable predictions (§25.3).

Real obstacles, clearly identified. The selectivity problem is not something critics discovered and enthusiasts hid. It is described in the field's own literature, quantified with a standard metric, and it is the explicit target of a large body of engineering work (§25.5). A field that names its own central problem in its own review articles is behaving well.

Real programs. Compounds have reached Phase 3. Companies have spent large sums. Trials have been run, reported, and — in the cases that failed — published (§25.7).

No verdict available. No approved systemic agent from this lineage exists as of this writing.

Now the part that separates an honest 🔬 from an evasive one.

A 🔬 that does not say what would resolve it is not a rating. It is a shrug. Anyone can say "more research is needed"; the phrase is so universally available that it carries no information. What makes 🔬 a genuine assessment rather than a refusal to assess is that it names, in advance and specifically, the observations that would move it.

This is the same discipline as falsifiability, applied to a rating rather than to a hypothesis. If you cannot describe what would change your mind, you have not formed a position — you have described your current comfort level. And the requirement cuts both ways: a 🔬 must also name what would move it down, or it is just optimism in a lab coat.

📊 Evidence Rating

Claim: Antimicrobial peptides will provide a new class of broad-spectrum systemic antibiotics for drug-resistant bacterial infections in humans. Rating: 🔬 Frontier (assessed as of this writing, 2026) Reason: The mechanism is real and well characterized; the obstacles are specific and known (selectivity margin, host-factor inactivation, manufacturing cost, and an antibiotic market that does not reward development); programs have run and continue to run; and no systemic agent from this lineage has been approved. There is no clinical verdict to report, and the absence is a statement about the field's stage, not about a claim someone has overreached on. What would change it: Upward — a randomized controlled trial of a systemic AMP-derived agent in a defined infection, showing efficacy at least comparable to standard care with an acceptable renal and hemolytic safety profile, would move this to ⚠️; approval plus consistent post-marketing use would move it to ✅. Downward — repeated Phase 3 failures for toxicity across structurally distinct candidates, or a demonstration that the selectivity margin cannot be widened enough in humans regardless of design, would move it toward ❌ for the systemic claim specifically, while leaving topical and immunomodulatory claims untouched.

Read the "what would change it" line again and notice four properties, because you will be asked to write lines like it in the exercises.

It is specific. Not "better trials" — a randomized trial, in a defined infection, with named safety endpoints.

It is bidirectional. It names what would raise the rating and what would lower it. A rating that can only go up is a hope.

It is bounded. It says which claim would move and which claims would not. Systemic failure does not falsify the topical claim, because they are different claims about different exposures.

It is dated. "As of this writing" is not a hedge; it is a timestamp, and it tells a future reader exactly what to re-check. If you read this chapter in 2032, the first thing to do is find out whether the trial described in that line has been run.

The failure modes on either side

Two ways to get 🔬 wrong, both common in writing about this field.

Using 🔬 as a compliment. "It's frontier science" can become a way of granting a compound credibility it has not earned, implying it is almost proven or on a track to approval. It is not a rank in a hierarchy that leads inevitably upward; most compounds at the frontier do not arrive. A 🔬 with a decade of failed programs behind it is a very different object from a 🔬 on a compound that entered trials last year, and the honest write-up says which one you are looking at. This chapter's is the first kind, and §25.7 says so.

Using 🔬 to avoid a ❌. The more insidious error. A compound sold to consumers today, with confident claims about human benefit and no human trials, is not at a frontier. The claim is ❌, whatever the state of the underlying science, because someone is asserting a conclusion the evidence does not carry. Interesting preclinical work never converts a confident unsupported claim into an open question. If a molecule is genuinely at the frontier and somebody is selling it as established, issue both ratings at once: 🔬 for the research program, ❌ for the marketing claim. Chapter 5's sixth rule makes that possible, and Part III uses it constantly.

📋 Your Evidence Dossier

Field 11 — Verdict — under genuine uncertainty.

You have been building Field 11 since Chapter 5: a short, dated, defensible summary of where a claim stands. Until now, the entries you have written have had answers to summarize. This chapter gives you the harder case, and it is the one that separates people who can assess evidence from people who can only report it.

How do you write a verdict when the honest answer is "too early"?

Both failure modes are tempting. The dismissal — "no approved drug after forty years, doesn't work" — is wrong on the evidence, since absence of an approval is not evidence of failure and §25.6 supplies a large non-scientific reason approvals may be missing. It also cannot be updated: someone who has concluded "doesn't work" has no reason to look again. The endorsement — "nature's own antibiotics, a revolutionary solution to the resistance crisis" — is wrong in the other direction, and its tell is that it contains no obstacle. A verdict on a field with forty years of unsuccessful development that does not say why it was unsuccessful is not a verdict.

The five components

1. State the claim precisely, with population and endpoint. Not "do AMPs work" — do AMP-derived agents, given systemically, treat bacterial infection in humans? Half of all apparent disagreements about this field dissolve once someone specifies systemic or topical.

2. Say what is established. Give the strong version honestly: mechanism demonstrated, ancient and near-universal biology, real breadth of activity in vitro, and approved peptide antibiotics already in clinical use.

3. Name the specific obstacle. Not "challenges remain." The selectivity margin, narrowed further by serum binding, salt sensitivity, and proteolysis; plus manufacturing cost; plus a market that does not reward the product.

4. Give the rating, dated. 🔬, as of the date you write it.

5. Name the resolving observation, in both directions — what would move it up, what would move it down, and which adjacent claims would be unaffected.

Worked demonstration

FIELD 11 — VERDICT                                    [worked demonstration]
COMPOUND: Antimicrobial peptides (as a class, systemic use)
DATE OF ASSESSMENT: [today's date]

CLAIM ASSESSED
  That AMP-derived agents, administered systemically, will treat bacterial
  infection in humans — specifically, drug-resistant infection.

RATING            🔬 Frontier

WHAT IS ESTABLISHED
  · Mechanism demonstrated biophysically; membrane disruption, not target binding
  · Near-universal across multicellular life; hundreds of millions of years of use
  · Broad in vitro activity, including against resistant organisms
  · Peptide antibiotics ARE approved and in use (colistin, daptomycin, vancomycin,
    bacitracin, polymyxin B) — the category is not speculative
  · Resistance appears harder and costlier to evolve than for target-specific drugs

WHAT IS NOT ESTABLISHED
  · That any host-defense-peptide-derived agent is safe and effective systemically
  · That the selectivity margin can be widened enough for intravenous use
  · That manufacturing cost can reach a level an antibiotic market can bear

THE SPECIFIC OBSTACLE
  Selectivity. Mammalian membranes are membranes. Discrimination is by charge and
  composition — a difference of degree, not kind. Margins narrow further in vivo
  through serum binding, physiological salt, and proteolysis. Murepavadin's Phase 3
  halt in 2019 for kidney injury is the concrete recent instance.
  Second obstacle, non-scientific and equally binding: antibiotic economics (§25.6).

WHAT WOULD MOVE THIS UP
  A randomized controlled trial of a systemic AMP-derived agent in a defined
  infection, efficacy comparable to standard care, acceptable renal and hemolytic
  safety → ⚠️. Approval plus consistent real-world use → ✅.

WHAT WOULD MOVE THIS DOWN
  Repeated Phase 3 failures for toxicity across structurally distinct candidates, or
  evidence that the margin cannot be widened in humans by any available strategy → ❌
  for the systemic claim.

WHAT THIS RATING DOES NOT COVER
  Topical AMP use (⚠️ — separate claim, lower selectivity bar)
  Colistin as last-line therapy (✅)
  Daptomycin for approved indications (✅)
  "AMPs cannot generate resistance" (❌)

CONFIDENCE IN THIS ASSESSMENT: high — the uncertainty is in the world, not in my
  reading of it. I am confident that the answer is not yet available.

That final line is the one most worth stealing. There is a difference between "I am uncertain because I don't know the literature" and "I am confident that the literature does not contain an answer." Both produce the words "we don't know," and they are opposite epistemic states. A verdict that distinguishes them is doing real work.

Your turn

For each peptide in your dossier, look at your Field 11 entries and ask: is any of them a 🔬 that should be a ❌, or a ❌ that should be a 🔬?

The specific test: is anyone confidently asserting a human benefit? If yes, and the human evidence is absent, the claim is ❌ no matter how promising the preclinical work — and you may need two entries for that compound, one for the research program and one for the marketing claim.

Then, for every 🔬 in your dossier, check that it names a resolving observation. If it does not, it is not a rating yet. Fix it before Chapter 26.

Conclusion

Antimicrobial peptides are an ancient, near-universal arm of innate immunity, made by essentially every multicellular organism, typically short, usually cationic, and characteristically amphipathic. Most work by disrupting bacterial membranes rather than by binding a specific molecular target, and that mechanistic difference is the source of the field's central hope: a bacterium can mutate a binding site far more easily than it can rebuild a membrane.

That hope is real and it is routinely overstated. Resistance to membrane-active peptides exists, takes several well-characterized forms, and — in the case of mcr — travels between organisms on a plasmid. The defensible claim is that resistance is harder and costlier to evolve, not that it cannot happen.

The obstacle that has actually stopped the field is selectivity. Bacterial and mammalian membranes differ in charge, cholesterol content, and transmembrane potential, and antimicrobial peptides exploit those differences — but they are differences of degree. At concentrations that reliably kill bacteria in a human body, many otherwise excellent peptides also damage human cells; the concentration needed in vivo rises above the assay number while the concentration tolerated does not. Layered on top is an economic structure in which the drug class we most need is the one least able to repay its own development, because good stewardship requires using a new antibiotic as little as possible.

And yet peptide antibiotics are in the hospital today. Colistin, abandoned for toxicity and recalled to service because the alternative became untreatable infection. Daptomycin, approved in 2003 and in routine use, defeated only by the lung. Vancomycin. Bacitracin in a bathroom cabinet. The accurate version of the field's failure is much narrower than the popular one: the specific project of turning innate immunity's host defense peptides into new broad-spectrum systemic antibiotics has not yet produced an approved drug.

That narrower statement is a 🔬 and not a ❌. ❌ is a judgment about a claim that outran its evidence; 🔬 is a description of a field doing things in the right order and not being finished. The difference is not politeness — it is a factual distinction about who has asserted what. And an honest 🔬 names what would resolve it. Here that is a randomized trial of a systemic AMP-derived agent showing efficacy comparable to standard care with an acceptable safety margin, plus manufacture at a cost some payer will bear. If you read this chapter years from now, that is the sentence to check. Chapter 26 turns to a very different corner of peptide medicine, where the evidence is settled and the argument is about something else entirely.

Key Terms

Antimicrobial peptide (AMP) — a short peptide, typically 12–50 residues, produced as part of innate immunity and capable of killing microorganisms. Also called a host defense peptide when its non-antimicrobial roles are emphasized.

Innate immunity — the fast, genetically encoded, non-specific arm of host defense that requires no prior exposure and generates no memory. AMPs are among its effectors.

Cationic — carrying a net positive charge at physiological pH, usually from lysine and arginine. The property that draws AMPs toward negatively charged bacterial surfaces.

Amphipathic — having spatially segregated hydrophobic and hydrophilic faces. In an α-helical AMP this arises from residues spaced three or four apart in sequence landing on the same side of the helix.

Defensin — a family of small, disulfide-stabilized, β-sheet AMPs. Humans make α-defensins (neutrophils, Paneth cells) and β-defensins (epithelia).

Cathelicidin — a family defined by a conserved precursor domain, stored inactive and released by proteolytic cleavage. Humans have one: LL-37.

LL-37 — the sole human cathelicidin; 37 residues, cationic and amphipathic, with antibacterial, LPS-neutralizing, chemotactic, and wound-repair activities and an implicated role in some inflammatory diseases.

Magainin — an α-helical AMP from Xenopus laevis skin, described in 1987; the compound class that opened AMPs to drug development.

Minimum inhibitory concentration (MIC) — the lowest concentration preventing visible bacterial growth under defined conditions; the standard measure of antibacterial potency.

Hemolysis — lysis of red blood cells; the standard in vitro readout of an AMP's toxicity toward mammalian membranes.

Selectivity index — the ratio of the concentration harming host cells to the concentration killing bacteria. The field's therapeutic window, and the metric on which most candidates fail.

Barrel-stave model — peptides assemble into a discrete transmembrane channel lined entirely by peptide.

Toroidal pore model — the membrane curves inward with the inserting peptides, so the pore is lined by both peptide and lipid headgroups.

Carpet model — peptides accumulate on the surface and destabilize the bilayer wholesale, detergent-like, with no discrete pore.

Lipopolysaccharide (LPS) — the major component of the Gram-negative outer leaflet; its lipid A anchor carries the phosphate groups that supply much of the surface's negative charge and serve as the polymyxin binding site.

Lipopeptide — a peptide bearing a covalently attached lipid tail. Colistin, polymyxin B, and daptomycin are cyclic lipopeptides.

Glycopeptide — a peptide bearing attached sugars. Vancomycin and teicoplanin are glycopeptides; their mechanism is cell wall inhibition, not membrane disruption.

Colistin (polymyxin E) — a cyclic lipopeptide that permeabilizes the Gram-negative outer membrane; abandoned in the 1970s–80s for nephrotoxicity and neurotoxicity, revived in the 2000s as a last-line agent against multidrug-resistant organisms.

Daptomycin — a calcium-dependent cyclic lipopeptide active against Gram-positives; approved in 2003, in routine use, inactivated by pulmonary surfactant and so not used for pneumonia.

mcr — plasmid-borne genes encoding enzymes that add phosphoethanolamine to lipid A, reducing outer-membrane negative charge and conferring transferable colistin resistance. First reported 2015.

Antimicrobial stewardship — restricting antibiotic use to preserve effectiveness. Clinically correct and commercially destructive, which is the core of the market problem.

Market entry reward — a proposed lump-sum payment on approval of a qualifying novel antibiotic, decoupling revenue from sales volume.

Subscription procurement — a payment model in which a health system pays a fixed sum for access to an antibiotic regardless of quantity used.

Spaced Review

1. (Ch 25) A news article reports that "antibiotic resistance kills five million people a year" and argues that antimicrobial peptides are the solution because bacteria cannot become resistant to them. Identify two separate errors, name the correct figure with its year and type, and state the defensible version of the resistance claim.

2. (Ch 25 + Ch 2) Chapter 2 described peptide hormones acting at picomolar concentrations through receptors that amplify a signal. Antimicrobial peptides typically act at micromolar concentrations and have no receptor at all. Explain what accounts for the roughly million-fold difference in required concentration, and then explain how that difference contributes directly to the selectivity problem of §25.5.

3. (Ch 25 + Ch 5) Colistin is rated ✅ in this chapter despite causing kidney injury in a substantial proportion of patients who receive it, while "AMPs cannot generate resistance" is rated ❌ even though the underlying peptides are genuinely promising. Using Chapter 5's rules, explain why both ratings are correct, and state in one sentence what a ✅ does and does not assert.

4. (Ch 25 + Ch 4) Chapter 4 covered why most peptides cannot be given orally and what determines where a peptide goes once administered. Using only that material plus §25.5, explain why topical antimicrobial peptide products have reached the market while systemic ones have not — and then explain why daptomycin, a systemic peptide, is nonetheless useless for pneumonia.

5. (Ch 25) Write a Field 11 verdict, in no more than 150 words, for the claim: "A defensin-based inhaled antibacterial will treat drug-resistant lung infection in people with cystic fibrosis." Assume preclinical work exists and no completed human trial has been published. Your verdict must include a rating, a date, one specific obstacle, and one bidirectional resolving observation. Then state, in one further sentence, whether an inhaled route makes the selectivity argument better or worse and why.