Introduction
Salmonella is a major foodborne pathogen, leading cause of non-typhoidal human salmonellosis. Eggs and egg products are most often implicated, although it can contaminate many food products. Fourier transform (FT-IR) spectroscopy allows for a rapid and cost-effective typing method of this bacterium at the subspecies levela, which is crucial for the early management of foodborne outbreaks. However, implementing it in a routine laboratory requires the development of a standardized protocol due to the high sensitivity of the method.
Materials and Methods
115 Salmonella isolates fully characterized by whole genome sequencing were included in the study. The strains, originating from food, feed and primary production, were collected during official controls by the FASFC in 2023. Of those 76 belonged to the serogroup O:4: Salmonella Typhimurium (n=19), S. Monophasic Typhimurium (n=19), S. Derby (n=18), S. Paratyphi B var Java (n=20); 15 to O:7 (S. Infantis) and 24 to O:9 (S. Enteritidis). The selected serotypes are under particular surveillance in the European Unionb. The portion of the spectrum studied reflects the carbohydrate composition of the outer membrane (1200 to 900 cm-1). Three biological replicates were tested for each single isolate, each of these was analyzed three times. Cluster analysis was done from the means of the spectra represented by scatter plots and dendrograms using principal component analysis (PCA) or linear discriminant analysis (LDA). Additionally, the classification capacity of the serogroups by the software's integrated classifier was evaluated.
Discussion
Cluster analysis revealed a discrimination power of 88,7% (102/115 isolates), of serogroups O:4, O:7, and O:9 using LDA. This can be explained by the selected spectral region, which reflects the carbohydrate composition, the main components of the somatic O-antigens determining the serogroup. Discriminating serotypes within serogroup O:4 would require more in-depth analysis to increase the performance. S. Infantis and S. Enteritidis, being the only representatives of the O:7 and O:9 serogroup respectively, do not allow for conclusions. The classifier's accuracy is 85.2% at O-group level (84,2% for O:4; 93,3% for O:7; 83,3% for O:9). Serotypes that displayed the higher rate of misclassification were S. Typhimurium serotype and its monophasic variant (both 26,3%). Whether through cluster analysis or serogroup determination by the classifier, 13 isolates were misclassified (11,3%). Further investigations are ongoing.
