Larazotide (AT-1001)

Identity, and a code number that belongs to three molecules

Larazotide is eight amino acids long: glycine, glycine, valine, leucine, valine, glutamine, proline, glycine. The FDA Global Substance Registration System records that sequence, GGVLVQPG, on both substance records — UNII ZN3R5560ZV for the free peptide and UNII FO8S2IW40N for the acetate salt. PubChem gives CID 9810532 for the free peptide, formula C32H55N9O10, molecular weight 725.8, CAS 258818-34-7; the acetate is CID 44146842, C34H59N9O12, 785.9, CAS 881851-50-9. KEGG drug entries D09351 and D09352 return the same two formulas and the same two CAS numbers. Vendor pages quote both weights interchangeably, which is a salt-versus-free-base distinction rather than a disagreement.

The sequence comes out of work on Vibrio cholerae. Di Pierro and colleagues published a structure-function dissection of zonula occludens toxin in 2001, built deletion mutants and tested them in Ussing chambers and receptor-binding assays. Within the processed carboxyl-terminal fragment they described an octapeptide receptor-binding domain, identified by amino acid comparison with zonulin and confirmed by site-directed mutagenesis. That octapeptide is the basis of larazotide. The paper reports a binding motif inferred from mutagenesis; it does not isolate or clone the receptor itself.

A second peptide from the same protein does the opposite thing, and the two are separated by one digit. AT1002 is the hexapeptide FCIGRL, mapped by Goldblum and colleagues in 2011 to zonula occludens toxin residues 288 to 293. In their work AT1002 reduced transepithelial electrical resistance across rat small intestine ex vivo, increased permeability to sugar tracers in vivo, and displaced ZO-1 and occludin from intercellular boundaries in a protease-activated receptor 2 dependent manner. AT1002 opens tight junctions and has been studied as an absorption enhancer. AT1001 was developed to close them.

Both major chemical databases carry a defect on this molecule. PubChem's synonym list for CID 44146842 includes "Zot protein, Vibrio cholerae (288-293)" — a six-residue span belonging to AT1002, attached to an eight-residue peptide. KEGG's D09352 lists a SEQUENCE of seven residues, dropping a valine, while the formula on the same record describes eight. The code itself is worse: a ClinicalTrials.gov intervention search for larazotide returns 29 studies, 19 of them Amicus Therapeutics trials of migalastat, a Fabry disease chaperone also coded AT1001. An unrelated nicotinic receptor antagonist described in 2012 was named AT-1001 too.

Zonulin, and what the assays actually measure

Larazotide is described almost everywhere as a zonulin antagonist or a blocker of zonulin receptors. Zonulin was identified as pre-haptoglobin 2, a human homologue of the cholera enterotoxin, and has been used since as a serum marker of intestinal permeability. Two things sit underneath that description and neither is settled. The receptor has never been molecularly identified: a PubMed search for the phrase "zonulin receptor" combined with terms for identification, cloning or characterisation returns one record, the 2001 Di Pierro paper, which infers a binding motif from mutagenesis rather than isolating a protein.

The measurement side has a documented problem. Scheffler and colleagues measured serum zonulin in 376 subjects from the Sorb cohort using a widely distributed commercial ELISA, found the values did not correlate with haptoglobin genotype, and then interrogated the kit directly with antibody capture, mass spectrometry and Western blot. None of the captured proteins corresponded to pre-haptoglobin 2. Recombinant pre-haptoglobin 2 added to the assay was not detected. The protein the kit did recognise was identified as properdin. Fasano, who first described zonulin, is a co-author on that paper.

Massier and colleagues extended the argument in a 2021 Gut letter, tabulating published correlations between commercial zonulin ELISA readings and direct intestinal permeability tests: R of 0.11 in 38 subjects, 0.17 in 71, and 0.033 in 24, none significant. Their letter also notes that pre-haptoglobin 2 is not naturally expressed in mice, which places a question over rodent studies reporting zonulin by ELISA. Sollid and Koning replied in the same journal under the title "Lack of relationship of AT1001 to zonulin and prehaptoglobin-2: clinical implications."

That reply could not be read for this page. It carries no abstract in PubMed and the publisher returned HTTP 403 to an automated request, so only its title, journal and status as a comment on the Massier letter are recorded here. The title's claim is the relevant one for anyone reading larazotide described as anti-zonulin: two immunologists who work on coeliac disease put in print that the relationship between the peptide and the protein it is named after is not established. None of the trials below measured whether larazotide bound zonulin.

The gluten-challenge trials

Paterson and colleagues ran the first human study in 2007, an inpatient double-blind randomised design in 21 adults with coeliac disease, 14 given a single 12 mg dose of AT-1001 and 7 given placebo, followed by acute gluten exposure. Permeability was measured as fractional urinary excretion of lactulose and mannitol. The placebo group showed a 70 percent increase after gluten; the treated group showed none. Interferon-gamma rose in 4 of 7 placebo participants and 4 of 14 treated ones, and gastrointestinal symptoms were recorded more frequently on placebo, P equals 0.018.

Leffler and colleagues published a dose-ranging study in 2012: 86 adults on a gluten-free diet, randomised to 0.25, 1, 4 or 8 mg three times daily or placebo, with or without a 2.4 gram daily gluten challenge for 14 days. The primary endpoint was the urinary lactulose-to-mannitol ratio, and it failed on its own terms. The authors reported the ratio was highly variable in the outpatient setting and that the gluten challenge itself did not raise it above the gluten-free control, which precluded assessment of any drug effect.

Kelly and colleagues reported the six-week version in 2013, in 184 patients on a gluten-free diet challenged with 2.7 grams of gluten daily while receiving 1, 4 or 8 mg three times daily or placebo. Permeability again showed no difference between any arm and placebo. Anti-tissue transglutaminase IgA told a different story: the mean ratio over baseline was 19.0 on placebo against 5.78, 3.88 and 7.72 at 1, 4 and 8 mg, with P values of 0.010, 0.005 and 0.025. Adverse event rates were similar across groups.

Across three trials the biomarker the whole programme was built on did not move. Hoilat and colleagues later pooled four randomised trials, 626 patients with 465 on larazotide and 161 on placebo, and found no significant difference in that ratio irrespective of gluten status. Symptom findings survived pooling in the gluten-challenge subgroup and did not survive it in the gluten-free diet subgroup.

The dose relationship runs backwards

Every trial that tested more than one dose found the lower one performed better. In Leffler's 2012 study the effect on symptom severity appeared at some lower doses and not at 8 mg. In Kelly's 2013 study the significant symptom result was at 1 mg, P equals 0.002, and the largest antibody suppression at 4 mg rather than 8 mg. The pattern then repeated in the largest trial of the programme, which used a dose range shifted downward to test it.

Leffler and colleagues randomised 342 adults with coeliac disease who had been on a gluten-free diet for at least twelve months and remained symptomatic, to 0.5, 1 or 2 mg three times daily or placebo. The design ran a four-week placebo run-in, twelve weeks of treatment and a four-week run-out. The primary endpoint, average on-treatment Coeliac Disease Gastrointestinal Symptom Rating Scale score, was met at 0.5 mg by both prespecified analyses, P equals 0.022 and P equals 0.005. The 1 and 2 mg arms did not differ from placebo on any endpoint.

Exploratory endpoints in the 0.5 mg arm moved in the same direction: a 26 percent decrease in symptomatic days, a 31 percent increase in improved symptom days, a halving or better of weekly average abdominal pain for six or more of twelve treatment weeks, and a decrease in headache and tiredness. The authors called the results mixed, and the description is accurate: a molecule that separates from placebo at 0.5 mg and not at 2 mg is either exhibiting a real non-monotonic relationship or producing findings that will not replicate.

No mechanistic account of the inversion has been published. A PubMed search combining larazotide with aggregation, self-inhibition, bell-shaped and inverted-U returned one record, and that record concerned migalastat in Fabry mice. Explanations circulate in secondary summaries — self-association at higher concentrations, receptor desensitisation, a saturable transport step — and none of them traces to a measurement made on this peptide in any system. The pattern is real in the published data and unexplained in the published data.

The phase 3 trial, and what it left behind

NCT03569007 was the registration study: a phase 3 randomised, double-blind, placebo-controlled trial of larazotide acetate for persistent symptoms in adults with coeliac disease on a gluten-free diet, sponsored by 9 Meters Biopharma, with a primary endpoint defined as the proportion of binary responders. It began on 29 May 2019. The registry records actual enrolment of 307, a status of TERMINATED, a primary completion date of 21 July 2022, and a stated reason of "Trial terminated by Sponsor." No results have been posted.

What the interim data showed is not in the public record in any form this page could verify. The registry gives no efficacy figures and no futility language. No publication reporting the trial appears in PubMed under any search combining larazotide with phase 3 or with coeliac disease, and an attempt to retrieve a contemporaneous patient-organisation summary returned HTTP 404. The trial that would have decided the question ran for three years, enrolled 307 people, and produced no retrievable data.

Larazotide has not been approved by any regulator for any indication. Ten of the 29 studies returned by a ClinicalTrials.gov intervention search are actually larazotide: eight sponsored by 9 Meters Biopharma, all in coeliac disease, and two by Massachusetts General Hospital. The commercial programme ended with the phase 3. What continued did so in an academic setting and in a different disease.

The post-COVID work

Yonker and colleagues published a mechanistic study in 2021 covering biospecimens from 100 children — 19 with multisystem inflammatory syndrome, 26 with acute COVID-19 and 55 controls — reporting prolonged gastrointestinal presence of SARS-CoV-2, raised zonulin, and spike antigen detectable in plasma. Appended to it was a single patient given larazotide, in whom plasma spike antigen and inflammatory markers fell alongside clinical improvement. One patient is one patient, and the authors framed it as proof of concept. Fasano, the senior author, is disclosed as a co-founder of and stockholder in Alba Therapeutics.

The trial that followed was small and did not finish as designed. NCT05022303, a phase 2a randomised, double-blind, placebo-controlled study at Massachusetts General Hospital, enrolled 12 children with a median age of 5.7 years, treated four times daily for three weeks with 24 weeks of safety follow-up, and was terminated with the registry reason "Decline in MISC cases." Yonker and colleagues published it in Science Translational Medicine in July 2025, reporting no larazotide-related adverse events.

Findings from that trial split into two kinds. The correlations are between antigen and illness rather than between drug and outcome: blood spike antigen tracked interferon-gamma at P equals 0.004, interleukin-6 at P below 0.0001, and gastrointestinal symptom score at P equals 0.003. The treatment comparisons are described directionally — faster resolution of symptoms, faster antigen clearance, faster return to usual activity — in twelve children, and the authors' own conclusion uses "may be safe" and "may improve". A sibling study, NCT05747534, a phase 2a in 107 children and adults with long COVID, is listed as active and not recruiting with completion due September 2026. Nothing outside coeliac disease and post-COVID syndromes has been registered anywhere.

Animal and cell work

The animal literature is small, recent and mostly from two groups in Malatya. Karahan and colleagues divided 32 male Sprague-Dawley rats into four groups and induced acute pancreatitis with intraperitoneal L-arginine, giving one group larazotide at 0.01 mg per mL in drinking water for seven days beforehand. Intestinal damage scores, ZO-1 immunoreactivity and serum FITC-dextran were lower in the treated pancreatitis group than the untreated one, all at P below 0.01, and bacterial translocation occurred in 50 percent of treated animals against 100 percent of untreated. Caliskan and colleagues used the same exposure in a thioacetamide liver-failure model, where intestinal damage scores fell but liver scores and serum ammonia did not.

Where the peptide goes after an oral dose has been measured once, in pigs. Enomoto and colleagues gave a 1 mg delayed-release oral dose, roughly 0.05 mg per kg, to overnight-fasted Yorkshire cross pigs and sampled intestinal fluid by in vivo ultrafiltration. Larazotide appeared in the distal duodenum and proximal jejunum at 0.32 to 1.76 micromolar at one hour and persisted at lower concentrations to four hours. Three pigs were used and two contributed to the reported analysis. That is the entire published in vivo disposition record for this compound, in any species.

Cell work continues to generate mechanistic proposals rather than converge on one. Kim and colleagues pretreated C2BBe1 and IPEC-J2 monolayers before anoxia and reoxygenation, reporting higher transepithelial resistance, preserved tight junction protein organisation and reduced myosin light chain-2 phosphorylation; the abstract states a pretreatment concentration of 10 mM, several orders of magnitude above anything measured in the pig study, and it is reproduced here as published. Glinka and MacGregor, in human keratinocyte monolayers, describe larazotide as a protease-activated receptor 2 antagonist — a different mechanism, in a different tissue.

What is not known

The measurement the programme was designed around never moved. Across three randomised gluten-challenge trials and a four-trial pooled analysis of 626 patients, the lactulose-to-mannitol permeability ratio did not differ between larazotide and placebo, and the 2012 study reported that the assay was too variable in the outpatient setting to support any conclusion. Whether the symptom findings reflect a barrier effect is therefore unestablished. No human pharmacokinetic data exist: no plasma concentration, no Cmax, no half-life for any species. No published stability, degradation or aqueous-solution time-course study for the peptide was located. The single phase 3 trial produced no public data at all, which means the largest dataset ever collected on this compound is unavailable and cannot be pooled with the phase 2 results. Zonulin was never measured as an outcome in any larazotide trial, and the commercial assay that would have measured it has been shown not to detect the protein it names. Nothing has been studied in pregnancy, in lactation, or beyond twelve weeks of continuous exposure in adults; the paediatric exposure record consists of twelve children treated for three weeks. Larazotide is not approved by any regulator in any jurisdiction, and material offered outside a registered trial has not been subject to the identity, purity or sterility controls that apply to investigational supply.

Questions