Methods are presented for investigating the site and form of growth of bacteria in model oil-in-water emulsions and in dairy cream. Following growth of the bacteria, the continuous aqueous phase is gelled using agarose and the oil phase removed using a mixture of chloroform and methanol. Using this method, the authors have found that Listeria monocytogenes, Salmonella typhimurium and Yersinia enterocolitica grow in the form of colonies in concentrated oil-in-water emulsions. Colonies of L. monocytogenes and Y. enterocolitica also form in artificially-inoculated fresh and tinned dairy cream. If information about the precise site of growth is not required, the authors have discovered that intact colonies can be liberated from the model emulsions by dissolving away the oil phase with chloroform: methanol.

The growth rates and yields of Listeria monocytogenes and Yersinia enterocolitica were determined in liquid culture media, and in model oil-in-water emulsions that contained 30, 70 or 83% (v/v) hexadecane. In emulsions with a mean droplet size of 2 pm containing 83% (v/v) hexadecane, the growth of both organisms resulted in decreased yields. Additionally, in these emulsions adjusted to pH 5.0 or 4.4 the growth rate of L. monocytogenes was significantly less than in other model systems which had an aqueous phase of equivalent chemical composition. Microscopic examination of the 83% (v/v) emulsion showed that its microstructure immobilized the bacteria, which were constrained to grow as colonies. Bacteria behaved similarly in model emulsions of either hexadecane or sunflower oil. Manipulation of the droplet size distribution of the emulsions changed the form and rate of growth of bacteria within them.

A predictive model was made for the logarithm of the thermal decimal reduction time (logD) of Salmonella enterica (D = time to 90% reduction by inactivation). The model was fitted with multiple linear regression from 521 logD-values reported in literature for laboratory media and foods highly varying in water activity and pH. The single regression model with temperature as the only variable had a high residual standard error (RSE) of 0.883 logD and no predictive value (fraction of variance explained (R²) < 0.001). Adding water activity, sugar content and pH as predictors resulted in a model with a lower RSE of 0.458 logD and an adjusted R² of 0.73. The model was validated by comparing 985 predicted with observed logD for S. enterica from other publications. The model was subsequently validated with 1498 published logD-values for inactivation of vegetative cells of nine other pathogenic bacteria genera (mainly Listeria monocytogenes, Escherichia coli, Clostridium perfringens, Cronobacter spp., Staphylococcus aureus, Yersinia enterocolitica) in or on a variety of laboratory media, meat, fish, dairy, nuts, fruits and vegetables. Regression analyses for validation with the 985 logD of S. enterica and 2483 logD of all genera show deviations from the expected slope of 1 (both 0.81) and the expected intercept of 0 (0.04 and 0.19 logD respectively). However, only 0.7% and 2% respectively of the new logD (expected: 0.5%) were observed above the 99% prediction interval of the original S. enterica model based on 521 logD. The findings suggest that i) the variability of thermal resistance of strains within species is larger than between genera and species; ii) one generic predictive model, also accounting for variability, suffices for designing the thermal inactivation of a variety of vegetative pathogenic bacteria in many food types.

A predictive model was made for the logarithm of the thermal decimal reduction time (logD) of Salmonella enterica (D = time to 90% reduction by inactivation). The model was fitted with multiple linear regression from 521 logD-values reported in literature for laboratory media and foods highly varying in water activity and pH. The single regression model with temperature as the only variable had a high residual standard error (RSE) of 0.883 logD and no predictive value (fraction of variance explained (R2) < 0.001). Adding water activity, sugar content and pH as predictors resulted in a model with a lower RSE of 0.458 logD and an adjusted R2 of 0.73. The model was validated by comparing 985 predicted with observed logD for S. enterica from other publications. The model was subsequently validated with 1498 published logD-values for inactivation of vegetative cells of nine other pathogenic bacteria genera (mainly Listeria monocytogenes, Escherichia coli, Clostridium perfringens, Cronobacter spp., Staphylococcus aureus, Yersinia enterocolitica) in or on a variety of laboratory media, meat, fish, dairy, nuts, fruits and vegetables. Regression analyses for validation with the 985 logD of S. enterica and 2483 logD of all genera show deviations from the expected slope of 1 (both 0.81) and the expected intercept of 0 (0.04 and 0.19 logD respectively). However, only 0.7% and 2% respectively of the new logD (expected: 0.5%) were observed above the 99% prediction interval of the original S. enterica model based on 521 logD. The findings suggest that i) the variability of thermal resistance of strains within species is larger than between genera and species; ii) one generic predictive model, also accounting for variability, suffices for designing the thermal inactivation of a variety of vegetative pathogenic bacteria in many food types.