Table_2_Phenotypic Prediction: Linking in vitro Virulence to the Genomics of 59 Salmonella enterica Strains.xlsx

The increased availability of whole-genome-sequencing techniques generates a wealth of DNA data on numerous organisms, including foodborne pathogens such as Salmonella. However, how these data can be used to improve microbial risk assessment and understanding of Salmonella epidemiology remains a challenge. The aim of this study was to assess variability in in vitro virulence and genetic characteristics between and within different serovars. The phenotypic behavior of 59 strains of 32 different Salmonella enterica serovars from animal, human and food origin was assessed in an in vitro gastro-intestinal tract (GIT) system and they were analyzed for the presence of 233 putative virulence genes as markers for phenotypic prediction. The probability of in vitro infection, P(inf), defined as the fraction of infectious cells passing from inoculation to host cell invasion at the last stage of the GIT system, was interpreted as the in vitro virulence. Results showed that the (average) P(inf) of Salmonella serovars ranged from 5.3E-05 (S. Kedougou) to 5.2E-01 (S. Typhimurium). In general, a higher P(inf) on serovar level corresponded to higher reported human incidence from epidemiological reporting data. Of the 233 virulence genes investigated, only 101 showed variability in presence/absence among the strains. In vitro P(inf) was found to be positively associated with the presence of specific plasmid related virulence genes (mig-5, pef, rck, and spv). However, not all serovars with a relatively high P(inf), > 1E-02, could be linked with these specific genes. Moreover, some outbreak related strains (S. Heidelberg and S. Thompson) did not reveal this association with P(inf). No clear association with in vitro virulence P(inf) was identified when grouping serovars with the same virulence gene profile (virulence plasmid, Typhoid toxin, peg operon and stk operon). This study shows that the in vitro P(inf) variation among individual strains from the same serovar is larger than that found between serovars. Therefore, ranking P(inf) of S. enterica on serovar level alone, or in combination with a serovar specific virulence gene profile, cannot be recommended. The attribution of single biological phenomena to individual strains or serovars is not sufficient to improve the hazard characterization for S. enterica. Future microbial risk assessments, including virulence gene profiles, require a systematic approach linked to epidemiological studies rather than revealing differences in characteristics on serovar level alone.