LL-37 (Cathelicidin): An Evidence Ledger

What the molecule is, and how it was named

In 1995 a group at Stockholm University screened a human bone-marrow cDNA library with a probe built from the pig gene PR-39, looking for a human counterpart to that peptide. Eight clones came back carrying something else. Agerberth and colleagues described a 39-residue peptide with no cysteine and no tryptophan, named FALL-39 after its first four residues, encoded by a 170-residue cathelin-like precursor and transcribed mainly in bone marrow and testis. The synthetic peptide was active against Escherichia coli and Bacillus megaterium in basal medium E. The name did not last. When Gudmundsson's group sequenced the gene and recovered the peptide from degranulated granulocytes, the mature species began two residues later and ran 37 amino acids.

That mature species is LL-37, and the identity data check out cleanly. The sequence is LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES, the formula C205H340N60O53, the molecular weight approximately 4493 g/mol, the CAS number 154947-66-7 and the PubChem identifier CID 16198951, with FDA UNII 3DD771JO2H recorded against the International Nonproprietary Name ropocamptide. hCAP18 is cleaved extracellularly to LL-37 by proteinase 3 following neutrophil exocytosis (Sørensen 2001, Blood 97:3951-9, PMID 11389039). Gudmundsson's 1996 gene paper concluded that this gene, then called FALL39 and now CAMP, is the only member of the cathelin family present in the human genome.

Homologues are a recurring source of confusion here, and one of them is embedded in the chemical databases themselves. The PubChem record for the human peptide carries, among its synonyms, a designation for the rabbit CAP18 fragment, which is a different molecule. Mouse work is usually done with CRAMP, which shares 67% sequence identity, and rabbit work with CAP-18. Overhage and colleagues showed in 2008 that the homologues are not interchangeable: at subinhibitory concentrations LL-37 inhibited biofilm formation and CRAMP did not. A rodent result reported under the heading of cathelicidin is therefore not automatically a result about LL-37.

The antibacterial figures, and the conditions attached to them

Turner and colleagues published the broadest early comparison in 1998, using radial diffusion and broth microdilution assays. They recorded minimum inhibitory concentrations below 10 micrograms per millilitre against Pseudomonas aeruginosa, Salmonella typhimurium, E. coli, Listeria monocytogenes, Staphylococcus epidermidis, Staphylococcus aureus and vancomycin-resistant enterococci, in media containing 100 mM sodium chloride. Three organisms behaved differently: methicillin-resistant S. aureus, Proteus mirabilis and Candida albicans were resistant at 100 mM salt but susceptible in low-salt media. Burkholderia cepacia resisted LL-37, protegrin and HNP-1 at either salt concentration. The same study reported high-affinity, positively cooperative binding to E. coli lipopolysaccharide, with a Hill coefficient of 2.02.

Johansson and colleagues, working in the same year on conformation, put a second set of numbers on the record and a limit beside them. LL-37 is disordered in water, adopts a helical structure on addition of bicarbonate, sulphate or trifluoroacetate, and loses helicity below pH 5. Helical content tracked antibacterial activity. The minimum inhibitory concentration against E. coli was 5 micromolar, and at 13 to 25 micromolar the peptide was cytotoxic to several eukaryotic cell types. In human serum, both the antibacterial and the cytotoxic activities were inhibited.

Overhage and colleagues reported a third figure that does not reconcile with the first. Against P. aeruginosa they recorded a minimum inhibitory concentration of 64 micrograms per millilitre, while biofilm formation was inhibited at 0.5 micrograms per millilitre through reduced attachment, stimulated twitching motility and downregulation of the Las and Rhl quorum-sensing systems. The three published figures do not convert into a single comparable series. Turner reported only an upper bound, below 10 micrograms per millilitre, which at this molecular weight is below roughly 2.2 micromolar; Johansson reported a measured 5 micromolar, and Overhage a measured 64 micrograms per millilitre, or approximately 14.2 micromolar. The published inhibitory concentrations therefore sit at or below approximately 14 micromolar, with one of the three unquantified at the low end, against different strains in different media by different methods. Averaging them into a single figure would produce a number that no experiment measured.

Biological fluid interferes further. Barańska-Rybak and colleagues tested LL-37 against S. aureus and E. coli in twelve human wound fluids, plasma fractions and serum, and found that five of the twelve acute wound fluids markedly inhibited its antibacterial action, with inhibition correlating with glycosaminoglycan content measured by Alcian Blue binding. Heparin-plasma strongly inhibited the effect. Cationic polymers that bind glycosaminoglycans restored activity.

The venous leg ulcer programme

The rationale for treating chronic wounds came from an observation about endogenous peptide, not from an antibacterial argument. Heilborn and colleagues reported in 2003 that hCAP18 levels in human skin rose after wounding, peaking at 48 hours and falling to pre-injury levels on closure, with the protein detected in the migrating epithelium over the wound bed. In chronic ulcers, levels were low and immunoreactivity was absent from ulcer-edge epithelium. In organ-cultured human skin, affinity-purified antibodies against LL-37 inhibited re-epithelialisation in a concentration-dependent way, and the proliferation marker Ki67 was absent from the epithelium of inhibited wounds.

Grönberg and colleagues ran the first-in-man study on that basis. Thirty-four participants with hard-to-heal venous leg ulcers went through a three-week open-label run-in on placebo, then a four-week randomised double-blind phase with twice-weekly application of LL-37 at 0.5, 1.6 or 3.2 mg/mL or placebo, then four weeks of follow-up. Healing rate constants were approximately six-fold and three-fold higher than placebo at 0.5 and 1.6 mg/mL, with p values of 0.003 and 0.088. Mean ulcer area fell 68% and 50% in those two groups. The highest concentration showed no difference from placebo, a non-monotonic pattern that secondary summaries of this trial rarely carry.

Mahlapuu and colleagues then ran the phase 2b. One hundred and forty-eight patients with hard-to-heal venous leg ulcers received 0.5 or 1.6 mg/mL or placebo alongside compression therapy. Mean age was 67.6 years, median ulcer duration 20.3 months and mean wound size at randomisation 11.6 cm squared. The efficacy analysis on the full study population identified no significant improvement over placebo. A post hoc analysis found improvement on several interrelated healing parameters in the subgroup whose target wounds were at least 10 cm squared at randomisation, and the authors describe that as an observation warranting a further, adequately powered study.

Both trials sit in the European registry rather than in ClinicalTrials.gov: EudraCT 2012-002100-41 and 2018-000536-10, both sponsored by the commercial developer of the topical formulation. A search of ClinicalTrials.gov alone therefore understates the human record for this peptide. No product containing LL-37 is approved anywhere; queries to the openFDA drug label and Drugs@FDA endpoints for ropocamptide and LL-37 return no match.

The rest of the human record

One registered trial has injected LL-37 into people. NCT02225366, at MD Anderson Cancer Center, was a single-group phase 1/2 study of intratumoural injection in melanoma, running from July 2015 to November 2020 with a registered enrolment of four. Results are posted. The participant flow lists three participants across two cohorts, one of whom withdrew for lack of efficacy. The primary outcome was the number of participants with an optimal biological dose defined by toxicity. No serious adverse events are posted; the other-events table lists eight terms, none affecting more than one participant in a cohort, including squamous cell carcinoma, actinic keratosis, anaemia, skin hypopigmentation and hypothyroidism.

A second registered study, NCT04098562, planned a quadruple-masked randomised comparison of an LL-37 cream against standard wound care in 40 patients with diabetic foot ulcers, sponsored by Fakultas Kedokteran Universitas Indonesia. Its estimated completion date was January 2020, its last registry update was September 2019, its status reads UNKNOWN and no results have been posted.

A separate strand of human work raises endogenous LL-37 rather than administering the peptide, and the distinction is routinely lost in summaries. Raqib and colleagues randomised 80 adults with shigellosis to a sodium butyrate enema or saline twice daily for three days; butyrate treatment induced LL-37 expression in the rectal epithelium and stool LL-37 remained higher on days 4 and 7, alongside earlier reduction of stool macrophages, IL-8 and IL-1 beta. Liu and colleagues had earlier shown that Toll-like receptor activation of human macrophages upregulated the vitamin D receptor and 1-hydroxylase, induced cathelicidin and killed intracellular Mycobacterium tuberculosis in culture.

The vitamin D question splits when tested in randomised trials, and the two results point in opposite directions. Quraishi and colleagues gave 30 septic ICU patients placebo, 200,000 IU or 400,000 IU cholecalciferol; median change in plasma LL-37 at day 5 was minus 17%, 4% and 30% respectively, p equal to 0.04, and change correlated with bioavailable but not total 25-hydroxyvitamin D. Raqib and colleagues randomised 129 pregnant women in Dhaka to weekly oral vitamin D3 or placebo from 26 weeks to delivery; supplementation did not raise LL-37 in plasma or in macrophage extracellular fluid, and stimulated intracellular LL-37 expression was significantly lower in the vitamin D group, p equal to 0.02, with no difference in bactericidal capacity.

Where an administered dose goes

Exogenous LL-37 has been tracked in one imaging study. Esposito and colleagues labelled an LL-37 variant with gallium-67 and followed it by SPECT/CT in healthy mice over 48 hours at a dose of just over 20 micrograms. Given as an intravenous bolus, the radiotracer cleared rapidly from blood, largely into the liver, with an appreciable fraction distributing temporarily to the lungs. Given subcutaneously at the same dose level, it was absorbed systemically following a two-phase kinetic model and cleared predominantly by the kidney. Uptake into lymph nodes and spleen was seen by both routes.

Two qualifications belong with that. A gallium-labelled variant is not the unmodified peptide, and the study describes distribution of the tracer rather than of biological activity. The species is mouse, and the mouse expresses CRAMP rather than LL-37, so the disposition environment is not the human one either. No human pharmacokinetic study of administered LL-37 was located, and the topical trials measured wound outcomes rather than exposure.

Searches of PubMed for a plasma half-life, clearance figure or serum degradation time-course for LL-37 returned no primary record. What the published chemistry does establish is that plasma is not an inert medium for this molecule: serum inhibits both its antibacterial and its cytotoxic activity in vitro, and glycosaminoglycan-rich fluids inhibit it further. Any half-life quoted for the peptide should be read against the absence of a study measuring one.

The same peptide in the pathology literature

Elevated cathelicidin is a feature of several disease states, which is the half of this literature that the antimicrobial framing tends to omit. Yamasaki and colleagues found abnormally high cathelicidin in the facial skin of people with rosacea, and forms processed differently from those in unaffected individuals, associated with increased stratum corneum tryptic enzyme. Injecting the rosacea-associated peptide forms into mouse skin increased inflammation, as did adding the protease and as did deleting the protease inhibitor gene Spink5; targeted deletion of Camp confirmed the peptide's role in the protease-mediated inflammation.

Lande and colleagues reported in 2014 that two-thirds of patients with moderate-to-severe plaque psoriasis harboured CD4-positive or CD8-positive T cells specific for LL-37, that those cells produced interferon gamma and Th17 cytokines, and that the frequency of circulating LL-37-specific T cells correlated with disease activity. That paper carries a 2015 corrigendum in the same journal; no retraction or expression of concern is attached to it or to any other source cited on this page.

Coffelt and colleagues reported that neutralising LL-37 in vivo significantly reduced engraftment of mesenchymal stromal cells into ovarian tumour xenografts in mice, inhibiting tumour growth and disrupting the fibrovascular network, and described the peptide as overexpressed in ovarian, breast and lung cancers. The registered melanoma study proceeds from the opposite premise, that intratumoural LL-37 stimulates an antitumour immune response. Both positions are represented in the review literature, which describes the role as context-dependent rather than settled.

Airway tissue supplies the finding most relevant to how the peptide is actually handled outside clinical settings. Thomas and colleagues challenged human nasal epithelial cells and mouse macrophages with LL-37 and recorded release of ATP, IL-6 and IL-8 along with cell death and lysis, associated with caspase-1 and caspase-8 activation rather than caspase-3/7, and blocked dose-dependently by a synthetic glycosaminoglycan. The authors' disclosed financial interest in that glycosaminoglycan is stated in the paper. The cytotoxicity finding is consistent with Johansson's 1998 measurement of eukaryotic cytotoxicity at 13 to 25 micromolar.

What is not known

The published record does not establish that administered LL-37 heals chronic wounds. The one trial that reported a healing difference enrolled 34 participants over four weeks of treatment and found its largest effect at its lowest concentration with no effect at its highest; the 148-patient phase 2b that followed found no significant difference in the full study population, and its supportive result is a post hoc subgroup that its own authors describe as needing a properly powered study. No human pharmacokinetic data of any kind exist: no plasma concentration-time curve, no half-life, no exposure-response relationship, no route comparison. The single biodistribution study used a gallium-labelled variant in mice, a species that expresses CRAMP rather than LL-37. Systemic administration in humans has never been studied; the only registered injection study was intratumoural, enrolled four patients across five years, and posted results for three. Nothing establishes a safety profile at any dose by any route beyond topical application in two trials and intratumoural injection in three people. The in vitro antibacterial figures do not agree with one another, sit at or below approximately 14 micromolar depending on strain, medium and method, with the lowest of the three reported only as an upper bound rather than a measured value, and are inhibited by human serum, by physiological salt for some organisms, and by glycosaminoglycan-rich wound fluid. Cytotoxicity to human cells has been recorded at concentrations close to the antibacterial ones. Whether the peptide promotes or restrains tumour growth is unresolved in the literature and the two positions are supported by different experimental systems. No preparation of LL-37 is approved as a medicine in any jurisdiction, and no published stability or handling characterisation for the peptide in solution was located.

Questions