Tracing the sources of human salmonellosis: A multi-model comparison of phenotyping and genotyping methods

Abstract

Salmonella source attribution is usually performed using frequency-matched models, such as the (modified) Dutch and Hald models, based on phenotyping data, i.e. serotyping, phage typing, and antimicrobial resistance profiling. However, for practical and economic reasons, genotyping methods such as Multi-locus Variable Number of Tandem Repeats Analysis (MLVA) are gradually replacing traditional phenotyping of salmonellas beyond the serovar level. As MLVA-based source attribution of human salmonellosis using frequency-matched models is problematic due to the high variability of the genetic targets investigated, other models need to be explored. Using a comprehensive data set from the Netherlands in 2005–2013, this study aimed at attributing sporadic and domestic cases of Salmonella Typhimurium/4,[5],12:i:- and Salmonella Enteritidis to four putative food-producing animal sources (pigs, cattle, broilers, and layers/eggs) using the modified Dutch and Hald models (based on sero/phage typing data) in comparison with a widely applied population genetics model – the asymmetric island model (AIM) – supplied with MLVA data. This allowed us to compare model outcomes and to corroborate whether MLVA-based Salmonella source attribution using the AIM is able to provide sound, comparable results. All three models provided very similar results, confirming once more that most S. Typhimurium/4,[5],12:i:- and S. Enteritidis cases are attributable to pigs and layers/eggs, respectively. We concluded that MLVA-based source attribution using the AIM is a feasible option, at least for S. Typhimurium/4,[5],12:i:- and S. Enteritidis. Enough information seems to be contained in the MLVA profiles to trace the sources of human salmonellosis even in presence of imperfect temporal overlap between human and source isolates. Besides Salmonella, the AIM might also be applicable to other pathogens that do not always comply to clonal models. This would add further value to current surveillance activities by performing source attribution using genotyping data that are being collected in a standardized fashion internationally.