Time to growth and inactivation of three STEC outbreak strains under conditions relevant for fermented sausages

Published models predicting growth and survival capabilities of shiga-toxin-producing Escherichia coli (STEC) under a(w) and lactic acid stress were validated by performing experiments with fermented sausage associated outbreak strains. Strain variation in inactivation and time to growth (TTG) were investigated for strains representing three serotypes (O103, O111, and O157). The TTG and growth boundaries of each strain were compared with predictions of a model for generic acid adapted E. coli and survival with predictions of two inactivation models. In addition, the influence of strain variation on the performance of the inactivation models, in terms of bias and accuracy factors, was illustrated. Strains with induced acid tolerance were used in broths containing 50 or 110 mM total lactic acid. The concentration of undissociated lactic acid (HLac) was adjusted by setting the pH-value, and water activity (0.900 to 0.995 depending on experiment) was adjusted by adding NaCl. The survival capabilities of the outbreak strains were good compared to the model predictions. The average bias factors of inactivation model predictions were within a factor of 2.2 depending on the strain used to validate the model indicating that inactivation rates of outbreak strains were slower than predicted. However, the observed rates were similar to the rates of a previously studied acid tolerant generic E. coli strain. Similarly, the time to growth of two of the strains (O103 and O157) was comparable with model predictions, whereas the growth capability of the third strain (O111) was lower than predicted. These results suggest that the properties of the most tolerant sausage outbreak strains are comparable to tolerant generic E. coli strains, which imply that suitable non-pathogenic E. coli strains are valid surrogates for fermented sausage outbreak strains. The relative sensitivity of strains depended on the environmental parameters and the response evaluated. The strain with the smallest log reduction at 20 °C was O157, whereas it was strain O103 at 8 °C. Under conditions unfavorable for growth, the time to growth was much shorter for strains O103 and O157 than for strain O111, whereas differences between strains were negligible under conditions favorable for growth. Depending on the response variable and the specific application the limitation of not addressing strain variation may lead to biased, fail-dangerous, predictions. Thus, solutions on how to best address strain variation in the development and validation of predictive models are needed.