The European Food Safety Authority (EFSA) asked the Panel on Biological Hazards (BIOHAZ) to deliver a scientific opinion on the evaluation of molecular typing methods for major food-borne microbiological hazards and their use for attribution modelling, outbreak investigation and scanning surveillance. In particular, this opinion addresses the first two terms of reference of the mandate, namely: (i) to review information on current and prospective (e.g. whole genome sequencing (WGS)) molecular characterisation and sub-typing methods for food-borne pathogens (e.g. Salmonella, Campylobacter, Shiga toxin-producing Escherichia coli (STEC) and Listeria) in terms of discriminatory capability, reproducibility, and capability for international harmonisation, and (ii) to review the appropriateness of use of the different food-borne pathogen sub-typing methodologies (including data analysis methods) for outbreak investigation, attribution modelling and the potential for early identification of food-borne organisms with epidemic potential and the integration of the resulting data in risk assessment.
In the approach taken by the BIOHAZ Panel to the reply to these two terms of reference, the starting point is a bacterial isolate from a human, food, animal or environmental source which has already been characterised to genus or species level. The BIOHAZ Panel acknowledged that in the future, bacterial identification and molecular typing may be combined in a single procedure and included in a culture-independent diagnostic process. There is very little relevant experience regarding the application of such metagenomic approaches in the food-borne zoonoses field and therefore this area is not considered in this Opinion.
The BIOHAZ Panel highlights that all bacteria are subject to genetic change (e.g. in response to environmental stress and human interventions such as antimicrobial or heavy metal use or vaccination), sometimes by mutation but more often by acquisition or loss of genetic elements. These changes can be followed by clonal expansion in the case of biologically successful organisms. Ongoing evolution driven by genetic change and selection has given rise to highly adaptable organisms that are able to exploit and expand into novel niches and extend their host range. Such evolution may also be linked to the emergence of various ‘epidemic’ strains of pathogens, such as Salmonella, in combination with other biological factors and epidemiological opportunities for dissemination. The molecular characteristics of organisms provide markers for investigation of outbreaks, attribution studies, and assessment of potential virulence or epidemic potential. The BIOHAZ Panel also points out that even with high-resolution molecular approaches, up to and including WGS analysis, it is not possible to establish how closely two isolates are related without an appreciation of the structure and diversity of the bacterial population in question. Further, to properly evaluate typing methodologies, data from strain characterisation should be linked with epidemiological metadata and the strain selection must be unbiased and statistically representative of the population to be assessed. International harmonisation of molecular characterisation outputs by means of standardisation or appropriate quality control procedures is essential. This includes controlling the accuracy of production of DNA sequences from WGS and the further interpretations of annotation pipelines.
For the evaluation of molecular typing methods, the BIOHAZ Panel established a set of pre-defined criteria based on the first term of reference. These criteria included: (i) discriminatory capacity (i.e. degree of discrimination between strains of different genotype), (ii) reproducibility and repeatability (i.e. consistency of results within and between laboratories, and over time), (iii) current international harmonisation (i.e. status with regard to availability and use of standard operational procedures, external quality assurance systems, harmonised nomenclature and data management tools), and (iv) the potential for future international harmonisation in situations where any of the sub-criteria under (iii) may not be currently harmonised.
Following the evaluation against those criteria, the BIOHAZ Panel concluded that molecular typing methods should ideally provide appropriate discriminatory power, reproducibility, capability for international harmonisation and reduced handling of and exposure to pathogens in the laboratories. No current typing method, whether phenotypic or molecular, complies with all these expectations. Several methods are often used in combination in order to obtain the resolution needed. The methods applied depend on the pathogen and on the application sought. These methods have proven track records of use, and for some of them (e.g. Multi locus sequence typing (MLST), Pulsed-field gel electrophoresis (PFGE)) extensive databases of valuable typing data have been collected. Further, methods based on WGS can replace and are increasingly replacing the numerous different methodologies currently in use in human and veterinary reference laboratories, and the same methods can be used for all organisms. An essential precondition is the availability of quality control methods, to ensure the reliability and consistency of molecular data generated, coupled with high quality bioinformatics support for the analysis of the data generated. The BIOHAZ Panel acknowledged that, regarding WGS, limited knowledge is available in relation to the technical errors that occur during sequencing and analysis and on the effect of genetic drift in the different bacterial populations over time, which may complicate the interpretation of results.
With regard to the review of the appropriateness of use of the different food-borne pathogen sub-typing methodologies for different food-safety related public health applications (i.e. detection and investigation of food-borne outbreaks of disease, food-borne source-attribution, early identification of food-borne organism with epidemic potential and their integration in risk assessment) the BIOHAZ Panel concluded that detection of outbreaks and their investigation in real-time would be enhanced by the generation of fully comparable molecular typing data from human, veterinary and food laboratories prior to submission to a central or connected databases. Some molecular typing methods (e.g. MLST, PFGE, Multi locus variable tandem repeat analysis (MLVA)) have been harmonised to a greater or lesser extent for the purpose of outbreak detection and investigation. The international development of harmonised platforms for WGS-generated data should be encouraged.
In relation to source-attribution analysis of food-borne pathogens, the Panel concluded that a major challenge of using data generated from molecular typing methods in source attribution models, in particular WGS data, will be to define meaningful subtypes providing an appropriate level of discrimination for source attribution. A high level of discrimination is not necessarily the best option. The applied method has to allow for some genetic diversity between isolates from human and animal/food sources, but only to the degree so that it can still be assumed that they originate from the same source. Independent of the choice of molecular typing method and approach for source attribution, it is important that the data included from human and potential sources are related in time and space. Source attribution analysis is, therefore, facilitated by integrated surveillance providing a collection of isolates from all (major) sources that should, to the extent possible, represent what the human population is exposed to.
In relation to the last of the applications, the BIOHAZ Panel concluded that the epidemic potential of a food-borne strain within a bacterial species, or even within a subtype varies considerably, and is a function of its inherent genetic characteristics and their expression combined with ecological factors including the opportunities to spread in the food chain. Prediction of the public health risk and epidemic potential of emerging strains of food-borne pathogens has not yet been possible. Nevertheless, if an epidemic strain has already emerged in a certain region such a strain can be rapidly characterised employing current molecular typing methods and thus serve to identify the occurrence of such strains in other regions for risk management purposes. High throughput WGS technologies offer new opportunities to characterise bacterial strains in great detail. The genetic information that these technologies provide will however need to be considered together with gene expression, host and ecological factors, including the opportunities to spread in the food chain. Finally, although there are differences between bacterial species, the principle of assessing the gene content in relation to fitness as a means to assess risk potential that has been used for the four organisms considered in this opinion should be applicable to any bacteria.
The BIOHAZ Panel makes a series of recommendations on important issues to be considered as these methods, in particular WGS analysis, have limitations when using the data they generate. Thus, modern molecular typing methods provide many opportunities for rapid and accurate determination of the genealogical relationships among bacterial isolates. Interpretation of the results generated by these methods for different public health applications requires this information to be placed in the context of the diversity, degree of genetic change (e.g. during storage of isolates or mutation during an outbreak and in reservoirs) and population structure of the particular pathogen in question. Therefore, large scale carefully co-ordinated studies are required to fully elucidate this. The development of more informative and easier to use bioinformatic tools for analysis of WGS data is needed. Multidisciplinary and integrated research programs are needed to develop and validate the use of detailed genetic information for ‘predictive’ hazard identification, accounting for gene expression and how this affects the fate of pathogens in the food chain and their interaction with human and animal hosts. Further recommendations are made on particular issues to aid the use of these methods and the data they generate for the different applications considered.