Human lactoferrin

Human and bovine lactoferrin are two different proteins doing the same job. Pierce and colleagues cloned the bovine sequence from a submaxillary gland library and put the figure on it in 1991: 69 percent identical to the human protein, and 64 percent identical to the mouse one. The human chain, sequenced by Metz-Boutigue and colleagues in 1984, runs 703 residues in two homologous domains, each holding one iron site and one sugar chain, at asparagine 137 and asparagine 490. The secreted bovine chain runs 689 residues and carries five potential glycosylation sites. Goodman and Schanbacher reached the same 689 from a different direction in the same year, sequencing bovine mammary mRNA and finding a 708-residue precursor behind a 19-residue signal peptide.

Those are small differences spread along a long molecule, and the distinction they mark keeps mattering. When lactoferrin is described as the iron-binding protein of milk and of the neutrophil, the protein meant is the human one. When a supply of the protein is needed in kilograms, cattle are the practical source. The largest trial recorded on this site, ELFIN, gave the bovine protein to 2,203 preterm infants at 150 mg/kg per day, and its result is on lactoferrin in preterm infants.

Getting hold of the human protein

Human milk is not a manufacturing feedstock, so the human protein reaches a trial as a recombinant copy. Two production routes have been published in detail, and both reports are explicit that scale was the problem being solved.

Ward and colleagues expressed human lactoferrin in the filamentous fungus Aspergillus awamori as a glucoamylase fusion, secreted into the growth medium and cut to the mature protein by the fungus's own KEX-2 peptidase. Yields passed 2 g/L. They reported that the protein still bound iron, still bound human enterocyte receptors and still killed a broad range of bacteria.

Van Berkel and colleagues took the other route and put the human gene into cattle, producing recombinant human lactoferrin in bovine milk at gram per liter concentrations. Their comparison with natural human milk lactoferrin is the useful part. Iron binding and iron release were identical. N-linked glycosylation differed, since the sugar chains come from the animal doing the expressing rather than from the gene. They then tested whether the difference showed up anywhere: in three mouse infection models, using both immunocompetent and leukocytopenic animals, the two proteins were equally effective and localized to sites of infection the same way.

So a recombinant human lactoferrin carries the human amino acid sequence with sugar chains from whichever organism made it. On the one occasion anyone compared the two directly, that difference changed nothing measurable in a mouse.

What it has done in people

Four randomized placebo-controlled trials have given recombinant human lactoferrin by mouth, all four using the same preparation, talactoferrin alfa. They cover three unrelated populations, two of them one trial each and severe sepsis twice, and they do not point the same way.

Both of the largest adult trials failed. FORTIS-M enrolled 742 patients across more than twenty countries, and the survival curves lay on top of each other: 7.49 months against 7.66, with a confidence interval straddling 1. The sepsis trial went worse. Its data safety monitoring board stopped it after 305 patients, for futility and for safety, and the authors' own conclusion is that oral talactoferrin did not reduce 28-day mortality "and may even be harmful". The gap was concentrated in patients with shock, where 33 of 105 on treatment died against 21 of 104 on placebo, P = 0.064.

Reading the sepsis result carefully matters. The primary endpoint, 28-day mortality, missed significance. The in-hospital and three-month figures cleared it. A trial stopped early produces exactly this pattern of ambiguity, and the honest summary is that it found no benefit and a mortality signal pointing the wrong way.

The registry trail of this program is worth following, because none of these four trials has summary results posted on its ClinicalTrials.gov record. FORTIS-M's record is marked completed, holds no results, and was last updated in August 2012, a year before the survival figures appeared in Annals of Oncology. A fifth talactoferrin trial, in relapsed lung and head and neck cancer, was terminated after four participants because, in its record's words, negative data from another trial led the sponsor to stop it. All of it is counted in the audit of registered lactoferrin clinical trials, where nine of 85 completed trials have posted a result.

What makes the sepsis pair worth separating from the rest of the table is that the smaller trial came first and was positive. Guntupalli 2013 gave the same 1.5 g dose by the same route to 194 adults and reported 28-day mortality of 14.4% against 26.9% on placebo, a relative reduction of 46.5%, at a two-sided P of 0.052 that the paper describes as meeting its protocol-specified endpoint. An editorial in the same issue of the journal asked in its title whether the finding was too good to be true. Both trials, that editorial and the reason an early-stopped trial can still carry a grade are set out on lactoferrin and sepsis in adults.

The preterm trial is a sixth the size of FORTIS-M and its authors set it up as a safety study with exploratory efficacy, which is how its infection result should be read. It sits alongside the much larger trials of bovine lactoferrin in the same population, which are covered on lactoferrin in preterm infants.

The same preparation has also been applied rather than swallowed. A phase 1/2 study in 2007 put talactoferrin alfa in a gel on diabetic neuropathic foot ulcers, and 46 of its 55 patients were randomized to a 2.5% gel, an 8.5% gel or placebo for twelve weeks. Ulcers shrinking by at least 75% ran at 47% and 53% against 25%, which came to P = 0.09 against a trial powered in advance for P below 0.1. Its own abstract reports that the study met its primary objective. It is the earliest study of this molecule in people, and it is reported in full, with the rest of the dermatology literature, under lactoferrin and skin.

What has not been tested

None of those three trials says anything about human lactoferrin as it occurs in milk. They tested a manufactured copy, by mouth, in sick people. The protein an infant swallows arrives at about 3.16 g/L in colostrum and 0.90 g/L in mature milk, figures recorded with the assay behind each on lactoferrin concentration, and whether that exposure does anything measurable is a different question from whether 3 g a day helps an adult with advanced lung cancer.

This page therefore carries no evidence grade. The scale used elsewhere on this site rates one claim at a time, and three trials in three unrelated populations cannot share a row: a null result in refractory lung cancer, a mortality signal in septic shock and a 50 percent infection reduction in 120 premature babies are three findings, and averaging them would destroy all three. Where a single claim about lactoferrin has been tested often enough to grade, it has its own page, and the benefits review collects them.

For the structural chemistry behind the sequence comparison above, including the 1989 crystallographic refinement and the four residues that hold each iron atom, see what lactoferrin is. For the functions attributed to the protein sorted by the preparation each was demonstrated in, see what lactoferrin does.