Andrew Watson, Yvonne Gunning, Mark Philo and Kate Kemsley of the Quadram Institute Bioscience describe a new method of testing for meat fraud based on differences in a key meat protein.
Europe’s horsemeat scandal of 2013 may seem like yesterday’s news but to Dutch horsemeat trader Willy Selten, jailed for two and a half years for his part in the fraud, it probably does not seem such a distant memory. Peter Boddy and David Moss were also convicted in 2015 for activities linked to ‘Horsegate’. More recently in October 2016 Alex Ostler-Beech from Hull and Ulrik Nielsen from Denmark admitted conspiracy to defraud linked to the sale of ‘beef’ containing horsemeat: they will be sentenced later this year following the trial of a third man, Andronicos Sideras from London, also charged with fraud linked to horse meat.
Events, such as Horsegate, – be it in the meat sector or other food areas – remain a constant threat. That threat continues to motivate scientists across the world to develop and refine a multitude of methods for detecting fraud of many kinds in all sorts of foods.
The method of detection depends very much on the task in hand. One of the most powerful methods, perhaps unsurprisingly, relies on testing for the presence of DNA of the unlabelled organism in the food, for example horse meat in a product labelled ‘beef’. DNA is, after all, the most direct measure for the presence of horsemeat. However, an alternative approach to species detection is to measure the sequences of proteins, which are dictated by the sequences in DNA.
an alternative approach to species detection is to measure the sequences of proteins'
Proteomics
One of the most widespread technologies for detecting proteins is mass spectrometry (MS). However, using MS to reliably detect and identify intact proteins is tricky. Instead, the protein of interest is digested with trypsin (enzymatic proteolysis) and the resulting peptide soup is passed through a liquid chromatography (LC) column to provide a dimension of peptide separation, then the peptides are presented to a mass spectrometer. The mass spectrometer will not only have to determine the masses of each peptide, but also try to determine its amino acid sequence. This it will do by smashing the peptides into fragments. Each of the many possible fragments arising from a population of peptides traversing the mass spectrometer will be a sub-string of the original peptide. The task then is to turn this vast amount of fragment mass data into information about sequences: this is achieved by database matching.
This proteomics route is used to explore a whole suite of proteins within an organism. In the sample, there might be hundreds of proteins and thousands of peptides, more perhaps than even the most expensive mass spectrometer can analyse completely; there is also the additional need for software and database searching. This requires a significant level of skill and investment. It is a promising route for achieving non-targeted testing for food adulteration – i.e. looking for the presence of ‘anything’ unexpected in the beef burger off the supermarket shelf - but when seeking horse adulteration in beef the full-blown proteomics approach is overkill.
Key proteins associated with the adulterating species, or rather the species-specific peptides arising from those proteins, can be used as species markers.'
Targeted testing
When there is prior knowledge of adulteration sources – for instance looking for horse in a sample, not just ‘anything unexpected’ – then it makes sense to use this knowledge in an intelligence-led approach. This is targeted testing. Key proteins associated with the adulterating species, or rather the species-specific peptides arising from those proteins, can be used as species markers. Now the challenge is more focused: look for just the chosen target peptides.
Modern mass spectrometers can be programmed to look for entities of specific masses and to ignore everything else, so time and laboratory resources are not wasted tracking peptides and fragments that are not of interest. Another bonus is that there is no database searching involved. If the target peptides are known in advance, then so too are the fragments and the experiment simply amounts to determining the intensities of anticipated fragments in the instrument’s detector.
Liquid ChromatographyMass Spectrometry
The ‘triple quad’ mass spectrometer is the perfect instrument for this kind of study. The ‘triple’ refers to the fact that there are three stages to the mass spectrometer. The first stage is a mass filter, which allows only peptide (ions) of preselected mass (charge to mass ratio) to pass. The second stage smashes those interesting peptides with inert gas atoms. This causes the peptides to break into fragments. Any one peptide will typically break in one place, usually the peptide bond between two amino acids, but with enough peptides passing through all possible fragments will be represented. Those fragments then enter the third and final stage, another mass filter, which scans over the fragments picking out those of predetermined mass. These are forwarded to an ion counter. To complete the picture, it is normal to introduce peptides via an LC column. This whole complicated process, from a peptide entering the first mass filter right through to selected fragments all being detected, takes just a few milliseconds.
This type of experiment, using preselected peptide ions and preselected fragment ions, is called ‘multiple-reaction-monitoring mass spectrometry’ or MRM-MS (Figure 1). A key piece of terminology is the combination of a specific peptide with one of its fragments. This is a transition, written as a pair of numbers - the two charge to mass ratios involved.
The ‘quad’ in triple quad refers to the set of four electrodes running parallel to the beam direction. A radiofrequency voltage applied across the electrodes provides the mass selection. There are thousands of triple quads or equivalent instruments in laboratories all across the world. They are used in all kinds of targeted testing, for example the routine drug residue testing in foods or pesticides in crops.
![]()
Figure 1 MRM detection of peptides
Meat authenticity
Several research groups have explored MS as a tool in food authenticity, particularly in meat, and several marker peptides have been documented. However, only a handful of research groups across the world have fully embraced MRM-MS for meat authenticity testing. These include Jens Brockmeyer’s group at the University of Stuttgart1,2, Kate Kemsley’s team at the Quadram Institute Bioscience in Norwich (formerly the Institute of Food Research) 3 and Stefano Sforza and his colleagues at the University of Parma.4
For detecting, say, horse in a ‘beef’ product using an MRM approach, the starting point is a suite of marker peptides. There are two obvious routes to populate that list. One is to use a discovery-class machine and a proteomics approach but focus on horse - this is the route taken by the Brockmeyer group. It allows the selection of some highly species-specific marker peptides which, coupled with the Brockmeyer group’s pioneering use of a fragments-of-fragments approach, gives impressive detection abilities. They have been able to detect adulterant pork or horse at around a quarter of 1% level, better than the industry standard, which deems 1% as the threshold for adulteration. They even detected horse in a supermarket canned corned beef product post-Horsegate.
Myoglobin
An alternative approach is to take a short cut and use a protein that is obviously present in the adulterant species. This is the route taken by our group in Norwich, which has selected myoglobin, the molecule that makes red meat red, as a target molecule. Both horse and beef contain myoglobin but the sequences differ by 18 amino acids scattered at various points across the molecule (Figure 2). The trick then is to pick peptides that are present in both species but contain one or more sequence differences. These give rise to different peptide masses and to different transitions, which are easily detected by MRM-MS. One advantage of this approach is that there is no discovery phase, only a decision on which molecule to use. Myoglobin is handy as a proof of principle molecule because it is plentiful, heat-stable, water soluble and widely studied. One down side is the need to be careful which peptides to work with since some are shared across many species.
Indeed, some species have identical myoglobin - that of horse is identical to the donkey and plains zebra. Bovine myoglobin is the same as that of American bison and yak. This means that, without further work, the adulteration of a beef burger with horse will give the same results as the adulteration of a yak burger with zebra, but this is not too much of a limitation.
![]()
Figure 2: Comparison of beef and horse myoglobin peptides
![]() |
| Liquid chromatographymass spectrometry |
Quantitation
There is a further advantage to using a protein, such as myoglobin, that appears in more than one species and it concerns quantitation. Knowing that a food contains unlabelled meat is a start, but the next question is invariably ‘how much is there?’. Quantifying the unlabelled component can determine whether it is adulteration (above 1%) or contamination (below 1%), but measuring quantities is much more challenging than simply noting the presence of the errant species. In principle, it means knowing how much of the target molecule is extracted from the sample in the first place and how efficient the MRM-MS system is at detecting the marker peptides. There are ways of doing these things but they are all laborious.
There is a further advantage to using a protein, such as myoglobin, that appears in more than one species and it concerns quantitation. Knowing that a food contains unlabelled meat is a start, but the next question is invariably ‘how much is there?’. If the unlabelled component is known to be adulteration – above 1% - as opposed to contamination, this gives a handle on the amount present, but determining quantities is much more challenging than simply noting the presence of the errant species. In principle, it means knowing how much of the target molecule is extracted from the sample in the first place and how efficient the MRM-MS system is at detecting the marker peptides. There are ways of doing these things but they are all laborious.
However, using ‘the same’ protein from two different meats opens the way to a quick and easy relative quantitation, which is what consumers, regulators and industry really want, for example 5% horse in beef. This is not absolute quantitation of the type ‘there is x milligrams of horse myoglobin in my beef burger extract’.
Beef and horse may both contain myoglobin, but since their amino acid sequences differ they are not strictly the same. Consequently, our group at Norwich describes them as ‘corresponding’. The same is true of some of the peptides arising from the digestion of the two myoglobins. Though some are identical, a few look similar but differ by a variant of an amino acid, so these are termed ‘corresponding peptides’.
The key idea is that the corresponding proteins will behave within their respective organisms in essentially the same way and that their extraction efficiencies will be the same so that the properties of corresponding peptides in the LC and MRM-MS will be comparable. If all this is true then simply calculating the ratios of intensities of matching transitions is a substitute for the relative levels of the two meats in the original sample. This is the so-called CPCP (corresponding proteins, corresponding peptides) approach. Proportionality must be maintained throughout for this to work: other detection technologies could be used in a similar way provided they too obey this basic requirement.
Using mixtures of four red meats - pork, lamb, beef and horse, our Norwich team has successfully demonstrated this type of relative quantitation. More recently we and other groups have moved beyond raw meats to consider cooked products and so-called complex foods, which comprise multiple ingredients and involve one or more stages of processing. The expectation is that MS-based methods will show their strength here since multiple-component foods can generate confusing DNA signals, whereas MS methods target known proteins. This could also be put to use for differentiating between different parts of an organism used in a food stuff, for example offal versus lean meat in ‘meat’ products.
More recently we and other groups have moved beyond raw meats to consider cooked products and so called complex foods'
![]() |
| Horse and beef lasagne – the recipe is prepared with different amounts of horse to allow accurate determination of horsemeat levels. |
Conclusions
It is worth bearing in mind that ‘protein’ means a lot more than just meat, though meat products are economically important and an obvious technology test bed. Most foods, apart from oils and fats, contains protein and given the incredible sensitivity of modern mass spectrometers, the options for detecting specific protein signatures via peptides, especially in a targeted mode, are very promising. As the Interpol-Europol series of Opson Reports5 demonstrates, food fraud of all kinds is a live issue. The list of target foods is constantly growing and the fraudsters are incredibly creative. Analytical scientists must continue to rise to the challenge.
Andrew Watson, Yvonne Gunning, Mark Philo and Kate Kemsley,
Quadram Institute Bioscience, Norwich Research Park, NR4 7UA
T: +44 (0)1603 255000 Email:andrew.watson@quadram.ac.ukWeb:www.quadram.ac.uk
References
1. von Bargen, C., Brockmeyer, J. & Humpf, H. U. Meat Authentication: A New HPLC-MS/MS Based Method for the Fast and Sensitive Detection of Horse and Pork in Highly Processed Food. J. Agric. Food Chem.62, 9428-9435, doi:10.1021/jf503468t (2014).
2. von Bargen, C., Dojahn, J., Waidelich, D., Humpf, H. U. & Brockmeyer, J. New Sensitive High-Performance Liquid Chromatography Tandem Mass Spectrometry Method for the Detection of Horse and Pork in Halal Beef. J. Agric. Food Chem.61, 11986-11994, doi:10.1021/jf404121b (2013).
3. Watson, A. D., Gunning, Y., Rigby, N. M., Philo, M. & Kemsley, E. K. Meat Authentication via Multiple Reaction Monitoring Mass Spectrometry of Myoglobin Peptides. Anal. Chem.87, 10315-10322, doi:10.1021/acs.analchem.5b02318 (2015).
4. Prandi, B. et al. Mass spectrometry quantification of beef and pork meat in highly processed food: Application on Bolognese sauce. Food Control74, 61-69, doi:10.1016/j.foodcont.2016.11.032 (2017).
5. Interpol. Operation OPSON V (public version) 2015. (2016).
