Mycotoxins: The Fungal Contaminants in Everyday Foods

Ask a mycotoxin analyst where the error in a result comes from and the answer is not the instrument. Modern methods can quantify these compounds at concentrations of a few parts per billion with good precision. The uncertainty in a mycotoxin result is dominated, overwhelmingly, by the step that happens before any laboratory is involved: deciding which handful of kernels out of a shipment weighing many tonnes will stand in for the whole thing.

This is not a minor technicality. Mycotoxin contamination is not distributed evenly through a lot. It concentrates in a small number of individual kernels or nuts, sometimes a very small number, each of which may carry a toxin load thousands of times the lot average. A shipment can comfortably exceed a regulatory limit on average while most randomly drawn small samples from it test clean.

Which Fungi Produce Which Toxins

Mycotoxins are secondary metabolites: compounds a fungus makes that are not required for its own growth, and which appear to serve competitive or defensive purposes. They are small, chemically stable molecules, and that stability is what makes them a food safety issue rather than a spoilage issue. Cooking, baking and roasting reduce some of them modestly and destroy few of them completely, and the fungus that produced a toxin can be entirely dead and gone while the toxin remains.

Three genera account for most of what is regulated. Aspergillus species produce the aflatoxins, the most closely regulated group and the one associated with liver damage and liver cancer on long exposure, along with ochratoxin A. Penicillium species also produce ochratoxin A, and patulin, which is chiefly an apple and apple juice problem. Fusarium species produce a wide family including deoxynivalenol, the fumonisins, and zearalenone.

The division that matters practically is between field fungi and storage fungi. Fusarium species infect the growing plant, so their toxins are largely determined before harvest by weather and by the crop’s own resistance. Aspergillus and Penicillium are primarily storage organisms that colonise grain after harvest when moisture and temperature allow, though Aspergillus also infects certain crops in the field under drought stress. This distinction determines who can do anything about a given toxin: field toxins are an agronomic problem, storage toxins are a logistics problem.

Crops and Conditions That Favour Growth

Fungal growth in stored commodities is governed less by absolute moisture content than by water activity, which measures how much of that water is chemically available rather than bound to starch and protein. Two grains at the same moisture percentage can have quite different water activities depending on composition, and it is the available water that a fungus can use.

Below a certain water activity, no fungus grows at all. Above it, growth begins slowly, and the storage fungi have thresholds low enough that grain which feels dry to the touch can still support them. Temperature interacts with this: the same water activity supports faster growth when warm. Toxin production has its own requirements, generally narrower than those for growth, which is why toxin levels do not track visible mould reliably.

Crop susceptibility follows a fairly consistent pattern. Maize is vulnerable to both aflatoxins and fumonisins, and is the crop where drought stress during grain fill most reliably precedes an aflatoxin problem, because stressed plants develop damaged kernels that Aspergillus colonises. Groundnuts and tree nuts are classic aflatoxin substrates, being oil-rich and often dried in the field. Wheat and barley are the deoxynivalenol crops, with wet weather during flowering being the dominant risk factor. Coffee, cocoa, dried vine fruit and wine are the usual ochratoxin concerns, and cereals contribute to ochratoxin intake as well.

Physical damage is a consistent amplifier across all of them. Insect feeding creates entry points and raises local moisture and temperature through the insects’ own metabolism. Mechanical cracking during harvest and handling does the same. This is why insect control and gentle handling reduce mycotoxin risk even though neither has anything obvious to do with fungi.

Why Uneven Distribution Breaks Sampling

Here is the statistical heart of the problem. Suppose a shipment of groundnuts contains a small number of heavily contaminated kernels among a very large number of clean ones. The average concentration across the whole lot might sit near a regulatory limit. Now draw a small sample at random. Most such samples will contain no contaminated kernel at all and will test as clean. A few will contain one, and will test extremely high.

The distribution of results from repeated sampling of such a lot is therefore not the tidy bell curve that underlies most measurement statistics. It is heavily skewed, with a long tail: many low results, occasional very high ones, and a mean pulled upward by the rare extremes. Ordinary intuitions about repeatability fail completely. Two honest laboratories testing two honest samples from the same lorry can differ by an order of magnitude without either doing anything wrong.

The severity depends on the commodity. The problem is worst where individual particles are large and contamination is concentrated in few of them, which describes tree nuts and groundnuts. It is less severe in finely divided commodities where each particle carries a smaller share of the total, and least severe in processed products such as flour, paste or oil, where processing has already homogenised the material. This is why a peanut butter sample is far more representative of its batch than a raw groundnut sample of comparable size.

Total measurement uncertainty in mycotoxin work is conventionally broken into three contributions: the variability of drawing the aggregate sample from the lot, the variability of subdividing that sample down to a test portion, and the variability of the analysis itself. Across a wide range of commodities and toxins, the first contribution dominates, often by a wide margin, and the analytical step contributes least. Investing in a better instrument while drawing a lazy sample improves the smallest term in the equation.

Sampling Plans for Bulk Commodities

Regulatory sampling plans are designed around the skew, and they look excessive until the reason is understood.

The core principle is that many small increments should be taken from many locations throughout the lot and combined into one large aggregate sample. Increments are drawn during movement wherever possible, from a flowing stream on a conveyor or during loading and unloading, because a moving stream can be cut across its full width and gives every part of the lot a chance of being represented. Sampling a stationary heap or a filled container instead means sampling only what is reachable, and contamination has no obligation to be reachable.

The aggregate sample for a large lot of a high-risk commodity is specified in kilograms rather than grams, and large lots are subdivided into sublots that are each sampled and tested separately. The final laboratory test portion may be only a few grams, so the sample must be reduced by a factor of many thousands without losing its representativeness.

That third row is where field practice most often fails. The entire aggregate sample must be ground, not a scoop from the top of it, and it must be ground finely, because coarse particles preserve the very unevenness the procedure exists to eliminate. Slurry milling, where the sample is ground with added water, improves homogeneity further for difficult commodities and has the additional benefit of controlling dust.

Subdivision must then be genuinely random. Scooping from a mixed pile is not, because particles segregate by size and density during pouring. A riffle splitter or rotary divider takes many thin alternating cuts and produces two portions that are statistically equivalent, and it is one of the least glamorous and most important pieces of equipment in the whole workflow.

Detection Methods From Strips to LC-MS

Once a properly prepared test portion exists, the analytical choice depends on whether the question is a rapid screening decision or a regulatory determination.

Lateral flow strips dominate intake screening at grain elevators and processing sites. The chemistry is a competitive immunoassay: the extract flows along a membrane carrying antibodies specific to the toxin, and toxin in the sample competes with a labelled reagent for antibody binding sites, so the test line becomes weaker as toxin concentration rises. A reader quantifies the line intensity, converting a visual judgement into a number. Results arrive in minutes, which is the entire point when a lorry is waiting at a weighbridge.

Strips have real limitations. Each is specific to one toxin or one closely related group, so screening for several toxins means running several tests. Antibodies cross-react to varying degrees with structurally similar compounds, which can inflate results. And the performance claimed for a strip applies to the matrix it was validated on, so applying a maize-validated strip to an unusual substrate is an assumption rather than a measurement.

Chromatographic methods provide the confirmatory answer. Extraction is followed by cleanup, historically using immunoaffinity columns that capture the target toxin on immobilised antibodies while everything else washes through. High performance liquid chromatography separates the extract, and detection is either by fluorescence, which suits the naturally fluorescent aflatoxins and ochratoxin, or by tandem mass spectrometry.

Mass spectrometry has become the default because it solves the multi-toxin problem. A single injection can quantify dozens of mycotoxins simultaneously, which matters because co-occurrence is the norm rather than the exception; a maize lot with fumonisins very often carries deoxynivalenol as well. Its principal difficulty is matrix effects, where co-extracted material from the food suppresses or enhances ionisation and distorts quantification. The defences are matrix-matched calibration standards and isotopically labelled internal standards that experience the same suppression as the analyte and cancel it out.

Regulatory Limits Across Regions

Mycotoxin limits are set nationally and regionally, and they differ enough that a consignment compliant in one market can be rejected by another. The variation is not arbitrary; it reflects genuine differences in dietary patterns, in the practical achievability of low limits given local agriculture, and in how conservative each authority chooses to be.

The general architecture is consistent. Limits are set per toxin and per commodity, with stricter values for foods eaten by infants and young children, whose intake per unit of body weight is higher and who are more vulnerable. Aflatoxin limits are typically expressed both for the most potent individual compound and for the sum of the group. Separate and much lower limits apply to the aflatoxin metabolite that appears in milk when dairy animals consume contaminated feed, which is why feed limits exist alongside food limits.

For a buyer, the practical implication is that a certificate of analysis is only as meaningful as the sampling behind it. A certificate stating a low result says nothing useful unless it also states the mass of the aggregate sample, the number of increments, and how the sample was reduced. Requesting that information is the single most effective piece of due diligence available, and it costs nothing.

Storage Practices That Prevent Growth

Testing detects a problem that has already occurred. Storage management prevents it, and the levers are few and well understood.

Drying to a safe moisture content before storage is the primary control, and it must be done promptly, because the hours immediately after harvest, when the crop is warm and wet, are when colonisation begins. The target moisture depends on the commodity and on how long it will be held, with longer storage requiring drier material. Oil-rich commodities such as groundnuts need lower moisture contents than cereals for equivalent safety, because water partitions differently in a fatty matrix.

Uniformity matters as much as the average. A bulk that averages a safe moisture content but contains a wet pocket will grow fungus in that pocket, and moisture migrates within a stored bulk driven by temperature gradients. Warm air rises through the centre of a bin, cools near the top, and deposits moisture there, creating a wet crust that spoils while the bulk beneath remains sound. Aeration to keep the bulk at a uniform temperature is the standard countermeasure, and it works by preventing the convection rather than by drying.

Temperature monitoring is the practical early warning. Fungal and insect metabolism generate heat, so a rising temperature in a monitored bin indicates biological activity long before anything is visible. Regular monitoring of stored bulk, with attention to trends rather than absolute values, catches problems while they are still localised.

Finally, sorting removes what prevention missed. Damaged, discoloured and shrivelled kernels carry a disproportionate share of the toxin load, so physical removal of the visibly defective fraction reduces average concentration substantially. Optical sorters that reject individual kernels on colour or fluorescence are used commercially for nuts and are among the few interventions that reduce contamination in an already-affected lot. They are not a licence to accept poor material, but they are a genuine tool.