Samuel Wrench and Mike Bromley review the analytical testing methods available to help the food industry detect contaminants in food and meet the recommendations of the Elliott Report.
Introduction
A conservative estimate of the annual cost of food fraud is around $15 Billion globally (1, 2) affecting 10% of all commercially sold food products (3). Food fraud is of course not a new concept. Historical and recent records are littered with examples where unscrupulous individuals have adulterated both high value (e.g. saffron and caviar) and mass produced foods (e.g. fish, coffee, wine) with low value substitutes to increase their profits. In many cases the additives are not fit for human consumption. The first law set down in British statute to combat food fraud was the Assisa panis et cervisiae (Assize of breadand ale) in the 13th Century. This was enacted to regulate the price, weight and quality of beer and bread. Since then numerous other laws have been implemented to address this issue including the recent European FIC Regulation, which outlines the rules for general food and nutrition labelling.
The recent horsemeat scandal is a prime example of the impact that food fraud can have on corporations and individuals. A year after the initial reports, sales of frozen ready meals were down 6% year on year (4), which led to the demise of several enterprises, such as Spanghero and Silvercrest Foods, along with associated job losses and a loss of trust in some major supermarkets and food producers, even those who were not directly involved in the crisis (5). A key question posed at the end of this crisis was how and why did it happen? Clearly the root cause was fraud but how did it become widespread and why did it go undetected for so long? The UK government commissioned a review, led by Prof. Chris Elliott, which addressed these questions; his report recommended that there should be zero tolerance of food fraud and that there needs to be a focus on intelligence gathering (6).
Current authenticity analysis
When undeclared horsemeat was first detected in processed foods in 2012, only two analytical methods were commonly in use in both commercial and regulatory laboratories to detect the presence of meat contaminants: assays based on Enzyme Linked Immunosorbent Assay (ELISA) or Polymerase Chain Reaction (PCR). As with many analytical tests neither technology is designed to detect meat directly, rather they detect specific factors within a matrix that are unique to a particular mammalian species. ELISA uses antibodies to detect specific proteins and PCR uses oligonucleotide primers to detect specific DNA sequences. Both assays are able to rapidly and reliably detect contamination of meat by surrogates in the majority of samples, however they work very differently and have some specific drawbacks (see Table 1).
On the whole most service laboratories that offer ELISA based analysis do so using kits developed by a small number of manufacturers. These kits, although fit for purpose, are somewhat limited in their capabilities. They are able to detect contaminants typically at a level of 1% but are only able to detect fixed and limited species (key contaminants, such as chicken, pork, beef, sheep and horse meats). These restrictions are predominantly as a result of the time and significant cost associated with identifying and validating antibodies that are both sensitive (so as to avoid false negatives) and specific (so as to avoid false positive results) and the return on investment that can be made from selling such kits. Furthermore, due to the way proteins are modified during processing, different antibodies and therefore different kits are required to detect both cooked and uncooked contaminants.
PCR assays are also offered by numerous kit manufacturers to detect key contaminants. However, analytical laboratories with active molecular biology research capabilities are able to develop and validate assays, with varying limits of detection (typically between 0.01 and 1%), and accredit to ISO17025 standards relatively rapidly (within weeks). The cost of development is relatively low (as primary analysis can be performed in silico); the costs of key reagents are negligible and most analytical laboratories will have quality control material at hand. This allows for a relatively rapid response, once a novel contaminant has been identified.
Despite the advantages afforded by PCR and the fact that several laboratories were offering horse contaminant detection prior to 2012, the assays were not performed. Why not? As horsemeat was rarely processed alongside most of the commercial meats consumed in the UK and it was not considered to be a cheaper meat, any risk based profile would have suggested that it was highly unlikely to be a contaminant in processed food. This, coupled with the relatively high costs of authenticity services (c. >£150 for detection of a single contaminant, with additional contaminants being charged at c. £50 per sample) resulted in only certain commonly processed meats being assessed. With hindsight more testing should have been done, however, given the large number of commonly eaten mammalian species globally, the cost of detecting all potential contaminants is vast and could not have been practically performed by either PCR or ELISA based methods.It is clear that the analytical service provision was inadequate to support the food industry in policing and preventing the sort of unforeseen food fraud exemplified by the horsemeat scandal. However novel diagnostics are now available to detect far larger numbers of contaminants in a cost effective manner.
Next Generation Sequencing: the future of authenticity analysis?
The most recent advance to be offered commercially is the detection of contaminants based Next Generation Sequencing (NGS) technology (also known as massively parallel sequencing). Sequencing technology has advanced significantly over the last 20 years. The announcement of the completion of the first human genome sequence came in 2003. It took around 13 years and cost $2.7 billion. In 2014, Illumina launched the Hiseq X Ten Sequencer with the capability of sequencing a human genome in 2-3 days at a cost of $1000 per genome and a throughput of 18,000 genomes a year. However this technology comes with a requirement for significant investment, with the hardware alone costing $10 million. Fortunately, in parallel Illumina and Life Technologies have adapted their sequencing technologies (the MiSeq and the Ion PGM respectively) to offer lower cost platforms that can be used to provide cost effective solutions for analytical laboratories.
NGS differs significantly from Sanger sequencing. With Sanger sequencing, a large amount of template DNA is needed, this is typically in the form of a PCR product. To get an accurate output, the input needs to be homogenous i.e. for a meat authenticity test, the DNA all needs come from the same species. So if a complex matrix is assessed, even one with only two different species, a sequencing reaction would fail to generate meaningful data. NGS technologies only require a single DNA molecule to derive an output and are termed ‘massively parallel’ as each single DNA molecule is sequenced independently in a ‘microreactor’. In this way very complex mixes of DNA samples can be assessed. This advance has been exploited by several research groups, which have developed methods of analysing and deconvoluting the complex nature of microbial communities (microbiomes) in clinical samples to identify the causative agents of disease and in foods to investigate the microbes responsible for various fermentation processes and spoilage (7).
Figure 1: Schematic representation of analysis of a sample containing 2 species by A: ELISA using a specific antibody (blue) that binds epitopes in the ‘blue’ species but not the ‘red’. The example shows detection by sandwich ELISA. B: PCR using oligonucleotides (blue) that can prime amplification from the ‘blue’ species but not the ‘red’. The example shows detection of double stranded DNA binding using a fluorescent dye (Sybr green) C:NGS amplicon sequencing using universal PCR oligonucleotides (black) that can prime amplification from all species. The amplified products are sequenced and the outputs are compared in silico to a database of sequences from several thousand species. In this way all contaminants (so long as their sequences are represented in the database) can be identified.
These assays, frequently termed ‘amplicon based metagenomics’ employ three key steps: PCR amplification (specifically amplification of a target region, such as mitochondrial 16s, 5s or 12s rRNA) followed by NGS and in silico comparison of each output sequence to a reference database (Figure 1). In contrast to standard PCR assays, which provide species specific results because oligonucleotide primers are designed to highly variable regions of a genome, NGS assays employ PCR to amplify a fragment of DNA using primers designed to regions of exceptionally high conservation (i.e. sequences common in all species) that flank regions with relatively low sequence conservation (i.e. vary significantly from species to species). PCR products are amplified from all species in a mixed sample and each DNA strand is independently sequenced. By comparing the output from the sequencer, which generates in excess of 8 million sequence reads even on the most basic platform, to an annotated database of sequences, one can discern the nature of a sample. Whereas in a microbiome sample there could be several hundred thousand different bacterial species or millions of fungal species, the number of mammalian or plant species is significantly lower making the analysis somewhat easier. An early attempt (8) to adapt this technology to the detection of contaminants in food showed some promise and now an NGS assay has been commercialised for the detection of in excess of 7000 different meat and plant contaminants (9).
Conclusion
The introduction of an NGS assay to detect several thousand biological contaminants simultaneously removes the need for the food industry to second guess what adulterant will be used by fraudsters and provides manufacturers, retailers and authorities with the ability to identify fraud that would have hitherto gone undetected. However significant challenges still face the analytical industry. For example, we are still unable to accurately quantitate the level of adulterants in food. NGS technologies have the potential to be adapted to provide true quantitative analysis, by performing deep sequencing without PCR amplification (i.e. true metagenomics analysis), however these assays are still at the development stage.
2 A. Kircher, NCFPD, Tools for Protecting the Nations Food Supply, June 5, 2012.
3 Estimated by the Food Standards Agency of the U.K., as reported by K. Everstine and A. Kircher, The Implications of Food Fraud, Food Quality & Safety magazine, June/July 2013. 3. https://www.fas.org/sgp/crs/misc/R43358.pdf)
