Food authenticity testing part 2: Analytical techniques

The sample is introduced into the mass spectrometer either directly (e.g. vapourised by a laser) or as the outflow of a separation techniques such as gas chromatography (GC-MS) or liquid chromatography (LC-MS). The molecules that are introduced are ionised (given an electric charge) and may be broken into charged fragments. They are then separated by the path or time they take to travel in a vacuum through an electromagnetic field, which in turn is dependent upon the mass/charge ratio (m/z) of each ion: in effect, dependent upon its mass. The mass of a molecule or of its characteristic fragments is a very selective characteristic on which to base identification, but is not completely unambiguous.

The spectrometer can either be pre-tuned to only allow fragments of a certain m/z to pass through, or continually scanned to detect all m/z s.  The former configuration is more sensitive, whilst the latter provides more identification information and is used in untargeted analytical approaches.

The scope of applications for mass spectrometry is highly dependent on the wide variety of possible operating mechanisms, modes and configuration. There are three key facets: the ionisation mechanism, ion separation mechanism, and resolution settings.

An electromagnetic field is used to either bend ions through a certain path (magnetic sector or quadrupole spectrometers), trap it for a certain length of time (Ion Trap spectrometers), or time its travel down an acceleration tube (Time of Flight, ToF, spectrometers). The electromagnetic fields in Trap and ToF spectrometers can be varied much faster than in quadrupoles, making them more suitable for full scan applications.

Multiple spectrometers can be used in series, to increase selectivity. They can be of the same, or different, separation mechanisms. Once the ion emerges from the first it is broken into smaller fragments, each of which then passes down the second spectrometer.  If two quadrupoles are used in series (called “triple quadrupole” or “MS-MS”) this gives the optimum sensitivity and selectivity for targeted analysis.  LC-MS-MS is the industry standard for detecting trace level adulterants such as melamine or Sudan dyes.  If a quadrupole is coupled with a fast-scanning spectrometer (Q-ToF or Q-Trap) this gives the potential to pre-select the molecular ion in the quadrupole then identify it by the full scan of its fragments (optimum selectivity). Modern Q-ToFs can be operated either as a triple quadrupole or as a Q-ToF, giving scope for dual-stage screening and confirmatory analysis.

Even more selectivity can be achieved by adding an Ion Mobility (IM) step within the series  (e.g. LC-Q-IM-ToF). This separates using the size and physical shape of the fragment, rather than on mass. Ion mobility detectors are familiar at airport security checks. It is very quick, and so dovetails well with the rapid electronics and computing needed for ToF full scan acquisition.

Mass spectrometry is transferrable technology, with test methods replicable from one laboratory to another. It is widely adopted and well embedded in laboratories. It has a wide scope of application, gives confidence in identification, and can be used quantitatively at very low detection levels. With full scan techniques there is scope for retrospective data-mining; looking at historical data for signals that were not sought or considered significant at the time.

Isotopes are forms of a chemical element that differ only by the number of neutrons in the nucleus. They have the same chemical properties, but different mass.  Isotopes can either be radio-isotopes (decay by emitting radiation) or stable (do not decay to any appreciable extent). Most elements are a mix of isotopes. Natural carbon, for example, consists of approximately 99% stable ¹²C, 1% stable ¹³C, a tiny amount of radioactive ¹⁴C., and even smaller amounts of other isotopes.  Because isotopes differ in mass, they can be measured by mass spectrometry.

The proportion of minor isotopes in the environment is specific to the geographic or geological region.  So, in principle, the ratio of major-to-minor isotopes could be used to tell the geographic origin of any food.  An example is the ratio of ²H (also called Deuterium, or D) to ¹H (hydrogen).  Seawater, H₂O, contains a very small natural proportion of D₂O.  As seawater evaporates and falls as rainfall the D₂O is naturally fractionated; because it is slightly heavier, the D₂O rainfall tends to fall before H₂O.  So a higher ratio of H/D builds up further inland, with a higher ration of D/H at the first mountain ranges that weather fronts tend to hit. 

Test methods to diagnose geographic origin usually measure the isotopic ratios of hydrogen, carbon, sulphur, oxygen, nitrogen and/or strontium, typically using multivariate analysis to seek statistical patterns that are indicative of a particular geographic region.

Stable isotopes ratios in food can also be indicative of production method differences. An example is the difference between mineral and organic fertilisers. Mineral fertilisers are synthesised from ammonia, which is ultimately manufactured from atmospheric nitrogen. This has a slightly different ¹⁴N/¹⁵N ratio to soil-based nitrogen, the source of most organic fertilisers.  So, in principle, the ratio of ¹⁴N/¹⁵N in a vegetable can be used to diagnose if it was organically fertilised.  In practice, this is difficult to interpret in real-world situations (see 3.3 – Limitations)

Isotopic differences can also be diagnostic of adulteration. For example, bees tend to forage on plants that photosynthesise using the C4 pathway, producing honey with a certain ¹³C/¹²C ratio. Cane sugar photosynthesises using the C3 pathway, which produces a different ¹³C/¹²C ratio. So adulteration of honey with cane sugar will change the ¹³C/¹²C ratio of the honey.

Climatic and geological regions do not conveniently match national boundaries. It would be unlikely, for example, to be able to differentiate Italian from Slovenian wine. The more granularity in origin that is sought, the more detailed and extensive the reference database of authentic samples must be. This can become the limiting factor.

If there are non-local inputs into the food production system (e.g. animal feed, fertilisers) then the reference database will need re-populating whenever these change in authentically-produced samples.

Natural variations in the isotope ratio being measured, due to specific and unpredictable local factors, can outweigh the difference being sought.

If there are complex inputs into a food production system (e.g. starter feed, medicated feed, drinking water, finishing feed) then it can be difficult to correlate a cause-and-effect relationship with the overall isotope ratio and the single input of interest.

Before any analysis, intact DNA must be extracted from the sample. Extraction conditions may need to be optimised for each sample type. There are a number of proprietary extraction reagents and systems, with Qiagen having dominated the market in recent years. A significant proportion of a DNA assay cost can be the licencing cost of the extraction reagents.

If universal primers are used then the section of amplified DNA should contain species-specific variations within the sequence of bases. The sequence of bases can be diagnosed by selectively including polymerase inhibitors (dideoxynucleotides) into the PCR mix. This is known as Sanger sequencing. Bases can then be separated by capillary electrophoresis, and compared with public databases. Such testing is used for species identification.

There are a number of alternative proprietary systems for diagnosing the sequences of bases which are much more rapid than Sanger sequencing. These are collectively known as Next Generation Sequencing. They include Solexa sequencing (Illumina), pyrosequencing (Roche) and SOLiD (ABI).

Specific primers are most commonly used for meat species identification.  Each primer pair is selective for one species. Combinations of different primers can be used within the same PCR assay; e.g. multiple different primer pairs for beef, horse, pork, goat, chicken, turkey, and donkey all within the same mix. This gives a highly selective presence/absence test for each species.

If a fluorescent dye is incorporated into the DNA during PCR then the progress of the amplification can be measured in real time. The more amplified DNA produced, the stronger the fluorescent signal. Monitoring the fluorescence signal vs cycle numbers gives a characteristic flat baseline, followed by steep increase, followed by plateau. The steep increase corresponds to the point where DNA copy numbers rise exponentially from too-few-to-detect to overly-saturated. The number of cycles before this steep curve occurs is indicative of the number of DNA copies in the original mix. It can be calibrated in a semi-quantitative manner. Alternately, the curve produced by a given species can be retrospectively calibrated as a percentage of the total primer-targeted DNA in a multi-species assay.

Rather than amplification by PCR, DNA can also be selectively detected by the use of hybridization probes. These are long strands of bases that have been engineered to form complementary strands with specific regions of target DNA in the sample, binding on like a strip of velcro. The probes are labelled to enable easy detection, for example with an antigen, a radiolabel, or a fluorescent tag. One common way of using the technique is to design a fluorescent tag that is quenched when the probe binds with its target. The mix is then slowly heated, and at a certain temperature the bound strands will separate and the probe fluoresce again. The plot of fluorescent signal against temperature (the “melting curve”) is characteristic of the strength of binding; it will be strongest when the bound DNA is an exact match for the probe (i.e. when the sample contains the target DNA).

Hybridisation probes are particularly suited for point-of-use applications, as the equipment is small and portable and the procedure is faster than PCR.

DNA measurements, even when semi-quantified in terms of DNA copy numbers, are extremely difficult to relate back to the g/kg concentration in the original sample (for example, g/kg adulteration of horsemeat in minced beef). The amount of DNA, particularly mitochondrial DNA, extracted from any sample is inherently variable. Different cuts of meat have different amounts of DNA. Fish is even worst – the amount of DNA that can be extracted from fish is so variable as to make quantitation meaningless.

Real-time PCR can be overly sensitive.  Coupled with the lack of g/kg quantitation, this means that an interpretative line cannot be drawn between low-level species substitution/adulteration and verification of factory cleaning / cross-contamination.

Not all food sample types have intact DNA that can be extracted. Highly processed meat products, stocks, soups and gelatins have very low amounts of viable DNA. Laboratories may need to take a try-it-and-see approach in these cases.

DNA barcoding can be applied to mixes of species, as even mixed sequences will give a database match. But if the species of interest is only present at a small proportion (approximately < 10%) of the mix then it is impossible to deconvolute and match to a database.

Public DNA databases are only populated by certified competent laboratories, but there are still occasional errors. This is as likely to relate to inconsistent taxonomy as to errors in the sequences themselves. Some of the reference sequences were generated from specimens of animals or plants collected in the 1950s and put into long-term storage. Verifying these can mean going back to original taxonomical drawings, which may have been misclassified.

The higher the resolution (stronger and higher frequency of magnetic field), the better the granularity in the spectrum, the more expensive the equipment and the more care needed in its environment and siting. Low-resolution instruments can be sited at point of use, but high-resolution instruments need well controlled laboratory environments.

High resolution SNIF-NMR is standardised and widely accepted for fruit juice authenticity, with a good robust reference database that has been built in collaboration with the juice industry. 

Low resolution NMR is ideal for on-line use in factories, giving results within seconds with little ongoing cost after the initial capital investment. It is well-accepted technology, having been used routinely for many years for measuring the approximate total fat percentage in meat raw materials or product, and so extending scope to applications such as distinguishing the species of frozen white blockfish by its total fat NMR spectrum is just an incremental change for some factories.

Spectroscopy exploits the fact that distinctive parts of a molecule will absorb energy at fixed wavelengths and then re-emit energy at discrete frequencies. These frequencies are shifted by neighbouring electrons in nearby parts of the molecule. In this respect, spectroscopy is similar to NMR. 

The energy is provided by shining light upon the molecule. Different light frequencies are absorbed by different types of molecular structure.

This is the most common use of infra-red. IR spectra are derived indirectly from the changes in an interference pattern of the incident and emitted light waves as the detector is moved; a statistical treatment called Fourier Transformation (“FT-IR”) which gives much improved sensitivity compared to conventional spectra.

Spectroscopy is suited to point of use and on-line applications; equipment can be small, portable, and give instantaneous results. NIR, particularly, is suitable for miniaturisation into handheld scanners and even smartphone applications. NIR light also passes through glass and thin plastic, meaning that products can be examined through packaging. All spectral imaging can be configured so that it is suited for unskilled operation, giving red light / green light results output. Compared to mass spectrometry and NMR, the equipment is cheap.