LL-37 Has a Human Dosing Record. I Found None of It Under the Skin
Three randomised trials I found put it on a wound and one injected it into tumours. The largest missed its primary endpoint, its sponsor went into liquidation in October 2023, and the dose response ran backwards twice, the second time in a different molecule.
What it is
LL-37 is the working end of the only cathelicidin gene humans have. Cathelicidins are antimicrobial peptides, short protein chains that puncture bacteria, each built with a pro-region called cathelin that gets cut away to release the active part. Gudmundsson and colleagues reported in 1996 that the human genome appears to carry only one of them, a gene called CAMP. (PMID 8681941) Its precursor is catalogued at UniProt as P49913, and residues 134 to 170 are LL-37:
LLGDFFRKSKEKIGKEFKRIVQRIKDFLRNLVPRTES
Thirty seven residues starting with two leucines, which is the whole of the name. Unlike the extracts elsewhere in this reference, that is one defined chain at a molecular weight of 4,493, so a laboratory can in principle confirm a vial holds it.
What the wound trials found
Grönberg and colleagues published the first-in-man trial in 2014: 34 people with venous leg ulcers, the slow healing sores that come of failing leg veins, given LL-37 solution at 0.5, 1.6 or 3.2 mg per mL or a dummy, twice a week for four weeks, randomised and double blind. Healing rate constants came out about sixfold and threefold higher than the dummy at 0.5 and 1.6, at P equal to 0.003 and P equal to 0.088, and only the first clears the conventional threshold. That was a first-in-man trial, and I could not reach its full text to establish which measure it had set as primary. (PMID 25041740) And the sentence I have not seen quoted anywhere in the material I read: no difference in healing was observed between the group on 3.2 mg per mL and the dummy. The highest concentration did nothing.
Mahlapuu and colleagues then ran the phase 2b, HEAL LL-37, in Poland and Sweden: 149 patients randomised and 148 treated at 0.5 or 1.6 mg per mL or a dummy, twice weekly for 13 weeks on top of compression bandaging. It missed. Confirmed complete wound closure, the primary endpoint, came out at 26.5 percent on 0.5 mg per mL and 24.7 percent on 1.6 against 25.3 percent on the dummy, across 144 patients. (PMID 34687253) What gets quoted instead is a subgroup whose wounds were at least 10 square centimetres, closing at 28.1 percent on the low concentration and 19.6 percent on the higher one, against 8.1 percent on the dummy, with only the low concentration reaching significance. The abstract itself calls that post hoc, meaning the subgroup was picked out after the results were in, and asks for a properly powered study I could find no record of. Those arms held 21, 21 and 24 people, so it is roughly six wounds closing against two. Promore Pharma AB entered voluntary liquidation on 12 October 2023.
The third trial, 25 people with diabetic foot ulcers in Jakarta, 13 on LL-37 cream and 12 on a placebo cream twice a week for four weeks, is quoted for a granulation index that favoured LL-37 at all four measurement days. Its registered outcomes, at NCT04098562, were healing rate, bacterial count and inflammation markers, every one of which failed to separate. Granulation index is not among them, and is called the primary measure only in the published discussion.
The one injection trial I found
In every search I ran, I found one trial that injected LL-37 into a human being. NCT02225366, run by the M.D. Anderson Cancer Center with the National Cancer Institute, injected it into melanoma deposits in the skin weekly for up to eight weeks, at 250 micrograms per tumour in the first group and 500 in the second, into two to four tumours a session, which works out at 500 to 1,000 micrograms a session in the first group and 1,000 to 2,000 in the second.
Read the enrolment before the result. Open 8 July 2015, complete 24 November 2020, four people enrolled. Results are posted for three, all aged 65 or over, all recorded as having an immune related complete or partial response, meaning the tumours went away or shrank on the criteria used for immune treatments, with one of the two in the lower dose group separately recorded as stopping for lack of efficacy, and no deaths and no serious adverse events. That response count is a secondary outcome measure, and the trial was single arm, so there is nothing to compare it against.
Whose evidence is it
LL-37 has a large literature, 2,520 PubMed records the day I searched, and almost none of it is about injecting some. Search the United States government trial registry for LL-37 as an intervention and 20 records come back, which sounds like a clinical programme; 18 of them are measuring LL-37 in saliva, gum fluid or blood while giving something else entirely, usually vitamin D but also butyrate, ivermectin, aminocaproic acid or a mouthwash.
The strongest clinical result in the whole file gave nobody LL-37. Mily and colleagues randomised 288 adults with sputum positive lung tuberculosis in Bangladesh, double blind and placebo controlled, to daily 5,000 international units of vitamin D3 by mouth, or 500 mg of phenylbutyrate twice daily, or both, or a dummy, alongside standard tuberculosis drugs. At week 4, 71 percent of 65 patients on the combination and 61.3 percent of 62 on vitamin D alone had cleared the bacterium from their sputum, against 42.2 percent of 64 on the dummy. That was one of two primary endpoints; the other, reduction in clinical symptoms at week 8, is not reported in the abstract I read. (PMID 26394045) Now read which arm did what. LL-37 rose measurably inside immune cells in the phenylbutyrate alone arm at week 12 against placebo, and that arm is not one of the two that hit the week 4 culture endpoint.
Which leaves the Lyme claim, the commonest thing this peptide is sold for. Sambri and colleagues in 2002 measured five cathelicidins against the bacteria behind syphilis and Lyme disease. For Borrelia, the Lyme bacterium, that abstract reports only that all five sat between 307 and 449.4 milligrams per litre and does not say which peptide sat where, and I could not reach the full text tables. For the related syphilis organism it does say, and the human peptide came last of the five. (PMID 12461010) Converted at 4,493, that Borrelia range is roughly 68 to 100 micromolar, three to eight times higher than the 13 to 25 micromolar at which LL-37 damages human cells in culture.
The dose response runs backwards
This is the most interesting thing in the file, and it happened twice independently. In the first-in-man venous ulcer trial, 34 patients, 0.5 mg per mL beat 1.6 and the highest concentration of 3.2 was indistinguishable from the dummy. In the range finding study for OP-145, an analogue built from a 24 residue window out of the middle of LL-37, four subjects at each of four concentrations, ear drops at 0.5 mg per mL were reported more effective than drops at 1 and 2, as a dose finding observation rather than a trial result, and 0.5 was carried into the 34 person trial that followed. A lower one, 0.25 mg per mL, was tested there and was not chosen. Two molecules, two tissues, two sponsors, both settling near the bottom of the range they tested.
It behaves like salt in a stew rather than like a painkiller. More is not more, and past a point more is worse.
There is an explanation available and it is not established. 0.5 mg per mL is about 111 micromolar, four to eight times the 13 to 25 micromolar at which Johansson and colleagues found LL-37 cytotoxic to human cells in 1998 (PMID 9452503), so the higher concentrations may be damaging the tissue they are meant to heal. I found no study that tested that, and consistency is not evidence.
Where the circulating dose came from
What circulates for injection is 100 to 250 micrograms a day under the skin, with advanced write-ups going to 500 micrograms a day, usually once daily and often five days a week, in blocks of two to four weeks. Reported here as convention, not as a recommendation. Two of the pages carrying those figures say where they came from, which is more than most manage: one that the numbers are anecdotal and unsupported by scientific data, the other that they come from supplier and community sources rather than clinical trials.
Where they did not come from is the melanoma trial, the only injected human dose I found published, and I found no page that cites it. What they look like instead is vial arithmetic, and that reading is mine: the common reconstitution instruction is two millilitres of bacteriostatic water into a five milligram vial, giving 2.5 mg per mL, which makes 100 micrograms come out at a round 0.04 mL and 250 micrograms at a round 0.10 mL.
Now the arithmetic that matters, also mine. Take 500 micrograms, a molecular weight of 4,493 and a nominal three litres of plasma, and ignore absorption and clearance entirely so the answer is a generous ceiling: about 37 nanomolar if every microgram reached the blood. The concentration that stops E. coli growing is 5 micromolar, roughly 135 times higher. Uninflamed tissue already carries 0.2 to 1.1 micromolar, six to thirty times more than the injection would add. The Borrelia range sits 1,800 times above it at the bottom and 2,700 at the top.
The route that was studied does the opposite. Its 0.5 mg per mL is that same 111 micromolar, put onto the wound bed at 12.5 micrograms per square centimetre, a figure the paper states rather than one I derived, so a whole application to the 11.6 square centimetre average wound in the phase 2b is about 145 micrograms. Roughly one day of the circulating injected dose, delivered onto a dozen square centimetres instead of into a whole body.
What the regulators have on file
I found no marketing approval from the United States or the European regulator, and none anywhere else in the searches I ran. The furthest I found it getting was the phase 2b that missed its primary endpoint, after which the sponsor liquidated.
The Federal Register is a clean negative I can stand behind, because the query is exhaustive against a complete corpus. Searching it for cathelicidin returns three documents and searching for LL-37 returns the same three, all National Institutes of Health notices of inventions available for licensing, from 2000 to 2002. I also read the full text of the notice announcing the July 2026 Pharmacy Compounding Advisory Committee meeting, which named seven peptides for discussion. The words cathelicidin and LL-37 do not appear in it.
The compounding history is second hand and I am labelling it that way. Two law firm summaries report that cathelicidin LL-37 sat in FDA's Category 2, the designation for substances the agency has evaluated and found to raise significant safety concerns in compounding, that it came off in April 2026 when the nominations were withdrawn, and that the committee will reconsider it before February 2027. Every fda.gov page I tried returned an error, so I could not confirm that from an FDA document.
What could go wrong
The topical record is unremarkable and I will say so plainly. Of the 207 patients treated across the three wound trials, about 135 received LL-37 rather than a dummy. In the 148 patient trial, where 97 got LL-37, severe local reactions affected five patients or fewer at any visit, and 12 non fatal serious adverse events occurred in 11 of the 148 patients treated, 7.4 percent, counted across the two LL-37 arms and the dummy arm together, none judged related to the drug. The other two reported no safety concerns and one mild irritant contact dermatitis in the thirteen who received LL-37.
The injection record is three people, with one published skin toxicity beside it. Dolkar and colleagues describe a 63 year old woman with stage 3C melanoma on weekly LL-37 injections who developed warty papules and blistering lesions after about 45 days, eleven of twelve biopsies showing atypical squamous growth resembling a fast growing skin tumour. It resolved within two months of stopping. She had been treated with nivolumab and with cisplatin, vinblastine and dacarbazine before the LL-37 course, which the case report states and which gives the eruption more than one candidate cause. (PMID 29665030) All three participants with posted results are recorded at 65 or over, so I cannot tell you whether this is one of them.
Then the part that never reaches a sales page. Lande and colleagues showed in Nature in 2007 that LL-37 binds the body's own DNA and condenses it into a form taken up by plasmacytoid dendritic cells, the immune cells whose job is to spot viral DNA, where it trips their alarm (PMID 17873860), and in 2014 the same group found two thirds of patients with moderate to severe plaque psoriasis carrying T cells that specifically target LL-37, getting into affected skin and tracking disease activity in blood (PMID 25470744). The same group found the pattern again in lupus in 2011: complexes of neutrophil antimicrobial peptides, not LL-37 specifically, bound to the body's own DNA in patients' blood, released as web-like structures by white blood cells, tripping the same alarm receptor, with patients making antibodies against both parts of the complex (PMID 21389263). Complexes of LL-37 and mitochondrial DNA are raised in human atherosclerotic plaque and plasma, and in mice an antibody against the mouse equivalent complex improved the lesions (PMID 26680206). In mice carrying human ovarian tumour grafts, neutralising LL-37 slowed the tumours' growth and disrupted the blood vessel network feeding them (PMID 19234121). Twice in animals, the treatment was subtraction rather than addition.
The counterweight gets the same weight. Kienhöfer and colleagues bred mice without cathelicidin, gave them the standard laboratory versions of lupus and arthritis, and found no difference from normal mice in autoantibodies, inflammatory signals, organ damage or arthritis severity. Antibodies against LL-37 ran higher in lupus patients than in healthy donors but did not track disease activity, which they read as a side effect of the peptide being massively overproduced rather than a cause of anything. (PMID 25535966)
Both halves need saying together. The record establishes that the body's own LL-37 takes part in the mechanism of psoriasis and is a target of the immune system in most patients with moderate to severe plaque disease. It does not establish that giving LL-37 to a person causes or worsens autoimmune disease, because I could not find any study that gave it to anyone with an autoimmune condition, and the one trial that injected it excluded by protocol anyone with active autoimmune disease requiring disease modifying therapy, anyone immunodeficient and anyone previously immunised with LL-37. That is an absence, not a reassurance. I could not find human data on whether giving it makes the immune system react against it either.
What no one seems to know
Whether it does anything at the injected concentration, and why the highest concentration failed twice in two tissues. I found no measurement of blood levels after a dose of LL-37 by any route, and no published investigation of the inversion.
Whether the large ulcer subgroup is real. It was post hoc, in a trial that missed its primary endpoint, the powered study its own authors asked for is one I could not find, and the sponsor is in liquidation.
How long any of it holds. The phase 2b ran 13 weeks of treatment then a four month observation period, poorly attended, and I found nothing following anyone beyond about seven months from randomisation.
What this adds up to
LL-37 is one of very few things in this reference where the molecule is real, the identity is checkable, and proper randomised trials exist. That deserves saying. But every one of those trials put it on a wound bed, and on the arithmetic above an injection under the skin lands two orders of magnitude below any concentration at which this peptide has been shown to do anything to a bacterium.
The direction of the evidence that does exist is the harder problem. In the atherosclerosis and ovarian tumour work, the intervention that helped was taking LL-37 away. In psoriasis it is the antigen. A seller pointing at a huge PubMed count is pointing at a body part rather than at a drug.
If you have bought LL-37, check whether the page you bought it from cited Sambri 2002 for Lyme disease, and if it did, whether it told you that the paper never says which of the five peptides sat where against Borrelia.
Frequently asked
Has LL-37 actually been given to people?
Yes. Three randomised placebo controlled trials that I found have applied it to a wound as a cream or a solution, treating 207 patients between them of whom about 135 received LL-37 rather than a dummy, and one trial injected it into melanoma deposits, with results posted for three patients. What I could not find, searching PubMed, the United States registry and the European register, is any human study giving it under the skin, into a muscle or into a vein.
Did the largest LL-37 trial work?
It missed its primary endpoint. HEAL LL-37 treated 148 patients with hard to heal venous leg ulcers across three arms, and confirmed complete wound closure came out at 26.5 percent on 0.5 mg per mL and 24.7 percent on 1.6, against 25.3 percent on the dummy. A subgroup with wounds of at least 10 square centimetres did separate, 28.1 percent on the low concentration and 19.6 percent on the higher one against 8.1 percent on the dummy, with only the low concentration reaching significance, but the paper's own abstract calls that analysis post hoc, the subgroup arms held 21, 21 and 24 people, and the adequately powered study the authors asked for is one I could not find a record of. The sponsor entered voluntary liquidation in October 2023.
Does LL-37 treat Lyme disease?
The paper that claim rests on measured five cathelicidin peptides against Borrelia, the Lyme bacterium, and its abstract reports only that all five needed between 307 and 449.4 milligrams per litre to stop it moving, without saying which peptide sat where. I could not reach the full text tables. Against the related syphilis organism the same paper does place them, and the human peptide came last of the five. That Borrelia range works out at roughly 68 to 100 micromolar, about three to eight times higher than the concentration at which LL-37 damages human cells in culture.
What dose of LL-37 do people use, and where did that number come from?
The figures that circulate for injection are 100 to 250 micrograms a day under the skin, with some write-ups going to 500, reported here as community convention and not as a recommendation. Two of the pages carrying those figures state that they come from anecdotal, supplier and community sources rather than clinical trials. I could not find a published human trial testing LL-37 under the skin at any dose. The concentrations I found studied are 0.5 to 3.2 mg per mL on a wound in leg ulcer patients, 0.5 mg per gram of cream in a diabetic foot ulcer trial, and 250 or 500 micrograms injected into each of two to four melanoma deposits.
Is LL-37 safe?
The topical record is unremarkable and the injection record is too small to support an answer. Across the three wound trials about 135 of the 207 treated patients received LL-37, and in the largest, 12 non fatal serious adverse events occurred in 11 of the 148 patients treated, 7.4 percent, counted across the LL-37 arms and the dummy arm together, none judged related to the drug. Three patients have posted results in the one injection trial I found. Separately, the LL-37 the body makes itself takes part in the mechanism of psoriasis and is a target of the immune system in two thirds of patients with moderate to severe plaque disease. I could not find any study that gave it to a person with an autoimmune condition, and the one injection trial excluded those patients by protocol. That is an absence rather than a reassurance.
Is LL-37 approved anywhere?
I found no marketing approval from the United States or the European regulator, and none anywhere else in the searches I ran. The furthest I found it reaching was a phase 2b trial in venous leg ulcers that missed its primary endpoint, after which the sponsor, Promore Pharma AB, entered voluntary liquidation on 12 October 2023 and its patent rights were moved into a subsidiary.
This content is for educational and informational purposes only and is not medical advice. Peptides discussed are research compounds and may not be approved for human use. Nothing here should be used to diagnose, treat, cure, or prevent any disease. Always consult a qualified healthcare professional before starting any peptide, supplement, or protocol. Individual responses vary. Do not self-administer compounds without proper medical supervision.
Get the next one in your inbox
Free. One email a week at most. Unsubscribe in one click.
Double opt-in · No spam · Unsubscribe anytime
More in Immune and thymic
- ARA-290 Has Real Randomised Trials, and Their Main Endpoint Was a Photograph of an EyeIts sequence does not appear in erythropoietin at all, which is the interesting part. Across a 1 mg to 8 mg range the response did not rise with the dose, the skin biopsy that diagnoses the disease did not move, and I found no trial registered anywhere after the 2017 announcement of progress.
- Thymalin Is Calf Thymus Extract, and Its Clinical Record Is Being Lent Forward to Two Other PeptidesThe registered active substance is the words thymus extract, 10 mg from the glands of calves and cattle under one year. Khavinson's group now says the active substances are the dipeptides KE and EW, sold separately as Vilon and Thymogen. In the geroprotection paper everyone cites, 24 people received Thymalin on its own.
- Thymosin Alpha-1: The Approval and the Evidence Are for Different ThingsThe only European registration I could verify is Italian, and its indication section is one sentence about influenza vaccination in immunocompromised patients. The two largest placebo controlled trials I found, in sepsis and in hepatitis C, both came back null.