Combined approach of time series and genomic analyses to characterize a Salmonella Goldcoast increase in the French poultry sector
Résumé
Salmonella is the second-most frequent cause of bacterial food poisoning in Europe. In this context, ANSES is involved in the food chain surveillance. In the frame of a laboratory network for food, animal and environment analysis, Anses centralises continuously serotyping results with epidemiological data associated to each strain. On this historical database, ANSES has implemented time series analysis able to early detect unusual increases in specific serotypes. At the end of 2018, an unusual increase of Salmonella enterica serotype Goldcoast (S. Goldcoast) was detected. Combination of statistical tool and genomic analysis provided informative characterization of this event and identified the hypothetical links between strains of human and non-human origin.
Material and methods: Time series analysis was based on three algorithms: Farrington, RKI (Robert Koch Institute) and Bayes. A signal is flagged if at least two algorithms generate a statistical alarm. A calculatory method adapted to low headcounts was applied in absence of statistical convergence of these algorithms. Strains collected by ANSES were serotyped and sequenced by Illumina technologies. Single locus polymorphism (SNP) analyses were realised by Snippy on genomes of non-human and human origin strains who were placed in the EnteroBase platform by the French National Reference Centre for E. coli, Shigella & Salmonella (NRC-ESS) at the Institut Pasteur. R18.0877 strain (Biosample SAMN11129919) downloaded from GenBank was used as reference.
Results: In October 2018, the first detection of the increase of S. Goldcoast in the food chain was realized by the low headcount method. In a second time, significant increases were relayed by direct application of the three algorithms from November 2018 to February 2019. Most of these strains (90%) come from the poultry sector and 78% of them were isolated in West-France (eight regions). SNP analysis highlighted that most of them were genetically close and constituted a cluster with a variability inferior to 10 SNP. The genomic analyses did not reveal link between strains of non-human and human origin.
Conclusion: The statistical tool has proved capacity to detect an unusual increase of one Salmonella serotype with a low representation in the food chain. The epidemiological data have shown that this event was mostly associated to the poultry sector from West-France. Poultry strains have a high genomic proximity that supposes a common source of contamination. Further analysis may be implemented to study potential genomic link between these poultry strains and S. Goldcoast isolated from other food and veterinary sectors.