Biotin, streptavidin, and the test that reads wrong

The cofactor, and the enzymes that carry it

Biotin is a bicyclic molecule — a fused ureido and tetrahydrothiophene ring, with a valeric acid side chain. Its biological job is narrow and well defined: it is ligated to a specific lysine residue on a biotin-carboxyl carrier domain, and it ferries a carboxyl group between the two active sites of a carboxylase enzyme. Zempleni, Wijeratne and Hassan surveyed the vitamin in BioFactors in 2009, and Said gave a longer account of its absorption and transport in Subcellular Biochemistry in 2012.

Tong reviewed the structures of the biotin-dependent carboxylases in Cellular and Molecular Life Sciences in 2013 — acetyl-CoA carboxylase, propionyl-CoA carboxylase, 3-methylcrotonyl-CoA carboxylase, pyruvate carboxylase and others — and described the crystal structures of several of the holoenzymes. The review also notes that deficiencies in the activities of propionyl-CoA carboxylase, 3-methylcrotonyl-CoA carboxylase and pyruvate carboxylase are linked to serious inherited disease in humans, which is the clinical context in which biotin biochemistry was originally worked out.

That is the whole of the cofactor story, and it is not where the interesting recent literature is.

Why diagnostics borrowed the molecule

Streptavidin, from Streptomyces avidinii, binds biotin with a dissociation constant in the femtomolar range — an affinity so high that the interaction is treated as effectively irreversible on the timescale of an assay. Weber, Ohlendorf, Wendoloski and Salemme reported the structural basis in Science in 1989: comparing refined crystal structures of apo-streptavidin and the streptavidin-biotin complex, they attributed the affinity to multiple hydrogen bonds and van der Waals contacts together with the ordering of surface polypeptide loops that bury the biotin inside the protein.

Diamandis and Christopoulos set out the diagnostic consequences in Clinical Chemistry in 1991. If an assay needs to tether an antibody to a solid phase, or to build a detection sandwich that survives washing, then a biotinylated antibody and a streptavidin-coated surface give a linkage that is specific, modular and essentially permanent. A large share of automated immunoassay platforms were built on exactly that architecture.

The design assumption is that the only biotin in the reaction is the biotin the manufacturer put there.

What free biotin in the sample does to the number

When a sample contains unbound biotin, it occupies streptavidin sites that the assay intended for its own biotinylated reagents. The direction of the resulting error depends on the format, and that is the part which makes the interference genuinely dangerous rather than merely noisy.

In a sandwich assay the analyte is bridged between a biotinylated capture antibody and a labelled detection antibody. Free biotin blocks the capture step, less signal is generated, and the reported result is falsely low. In a competitive assay a biotinylated analogue of the analyte competes with the analyte itself for a limited number of sites; free biotin displaces the labelled competitor, and the reported result is falsely high. One interferent therefore pushes different analytes in opposite directions within the same sample.

Trambas, Lu, Yen and Sikaris characterised this systematically in Annals of Clinical Biochemistry in 2018. They titrated known concentrations of biotin into serum and measured the effect across Roche Elecsys assays, choosing a concentration range intended to span what has been reported both in people taking ordinary supplements and in high-dose therapeutic settings. They reported high variability in tolerance: troponin T, TSH and antithyroid antibody assays were affected at the lowest concentrations examined, 15.6 and 31.3 ng/mL, while the majority of assays were relatively resistant until higher ones. At 500 ng/mL and above, every assay in the panel showed significant interference, with the magnitude still varying by assay.

The authors make a methodological point worth keeping. The shape of the dose-response curve carries more information than the single tolerance figure a manufacturer quotes, because two assays with the same nominal threshold can fall off at very different rates beyond it.

A hormone panel that looked like a disease

Piketty and colleagues published the clearest demonstration of what this looks like clinically, in Clinical Chemistry and Laboratory Medicine in 2017. Working with 23 plasmas from multiple-sclerosis patients and healthy volunteers receiving high-dose biotin, plus 39 unsupplemented patients, they measured free T3, free T4, TSH, parathyroid hormone, 25-hydroxyvitamin D, FSH, LH, prolactin, C-peptide and cortisol on streptavidin-biotin platforms, and quantified biotin and its main metabolites by liquid chromatography tandem mass spectrometry.

Biotin concentrations in the high-biotin plasmas ranged from 31.7 to 1160 micrograms per litre. Most of the hormone results in those samples were abnormal — the authors note that the pattern can mimic various endocrine disorders. They then applied a procedure using streptavidin-coated microparticles to strip biotin out of the sample before assay, after which biotin was below the detection limit in all but two samples and the hormone results normalised. Re-assaying the same high-biotin plasmas on methods that do not employ streptavidin-biotin binding gave normal results, apart from two slight parathyroid hormone elevations. In the 39 unsupplemented patients the stripping procedure changed nothing, apart from an 8.4% decrease in parathyroid hormone.

That is a well-constructed experiment. The artefact was reproduced, removed by a chemical intervention aimed specifically at the suspected cause, and independently absent on an orthogonal method. It is the reason the interference is no longer treated as anecdotal.

How often it is actually present

An interference matters in proportion to how often the interferent is in the tube. Katzman, Lueke, Donato, Jaffe and Baumann addressed that at the Mayo Clinic and reported two parallel measurements in Clinical Biochemistry in 2018.

They distributed 4000 questionnaires to outpatients over one week in July 2017; 1944 were returned, a 48.6% response, with a median age of 62 years. Of those respondents, 7.7% (95% CI 6.6 to 8.9%) indicated biotin use. Separately, they quantified biotin in 1442 residual waste plasma samples from patients presenting to the emergency department, drawn from physician-ordered electrolyte panels, and found that 7.4% (95% CI 6.2 to 8.9%) were at or above 10 ng/mL — the lowest known interference threshold for the Roche immunoassays in question.

Roughly one sample in thirteen arriving at that emergency department carried enough biotin to put a susceptible assay outside its stated tolerance. The authors frame this as identifying a population at risk rather than as an estimate of harm, which is the correct reading: the study measured exposure, not misdiagnosis.

The regulator's position, and where it has not been resolved

The US Food and Drug Administration issued a safety communication on biotin interference with laboratory tests in 2017 and has since updated it. Its stated concern is specific: the agency says it has continued to receive adverse-event reports indicating that biotin interference caused falsely low troponin results, and it publishes a list of listed troponin in vitro diagnostic devices that are subject to biotin interference and whose manufacturers have not addressed the risk. A falsely low troponin is the failure mode with the least margin for error, because that number is being used to decide whether cardiac injury has occurred.

The agency also issued guidance for industry and its own staff on how biotin interference should be tested for in in vitro diagnostic devices, which moved the problem from a laboratory-medicine discussion into a device-review requirement. Some manufacturers have since raised the biotin tolerance of individual assays, which is why a tolerance figure quoted in a paper from 2017 or 2018 does not necessarily describe the current version of the same assay.

All of the above is a measurement problem rather than a toxicology one. None of it describes a physiological effect of biotin at any intake, and none of it is a protocol, a schedule, or guidance on use for anybody.

The rest of the B group, and the one with a real toxicity record

The combined B-group fill in this catalog is a different object from a single cofactor, and it is worth being explicit about which parts of the B-group literature are solid. Pyridoxine is the clearest cautionary case, and the human record there is bad rather than merely thin. Schaumburg, Kaplan, Windebank, Vick, Rasmus and colleagues described a sensory neuropathy in people who had taken large amounts of pyridoxine, in The New England Journal of Medicine in 1983, titling it a new megavitamin syndrome. It remains the standard reference for the fact that a water-soluble vitamin is not thereby harmless, and the neuropathy it describes is a documented human injury rather than a theoretical risk.

Vitamin B12 contributes a measurement problem of its own, structurally similar to the biotin one in that the assay rather than the biology is the hard part. Hannibal, Lysne, Bjørke-Monsen, Behringer, Grünert and colleagues reviewed the biomarkers and decision algorithms in Frontiers in Molecular Biosciences in 2016, and the reason that review exists is that total serum B12 is a poor standalone indicator; the functional markers and the algorithms built on them are attempts to work around it.

The useful generalisation across both: for several compounds in this group the most consequential published literature is analytical. What a laboratory can reliably measure is the limiting factor on what anybody can conclude.

The limits

Neither biotin nor the combined B-group fill in this catalog is supplied for use in a person, and nothing above should be read as describing what either would do in one. The biotin literature summarised here is enzymology, crystallography, serum-spiking experiments on diagnostic instruments, a hormone-panel study in supplemented plasma, and one prevalence survey at a single centre.

The reason this article sits on a research-supply site is narrow and real: anyone running immunoassays in a laboratory that also handles biotin has a contamination pathway to think about, and the published tolerance figures are the right place to start. The certificate-of-analysis article covers what a lot document does and does not establish, and the bench-handling article covers the adjacent reconstitution and storage questions.