Effects of Temperature, ph, Glucose, and Citric Acid on the Inactivation of Salmonella typhimurium in Reduced Calorie Mayonnaise

1497 Journal of Food Protection, Vol. 60, No. 12, 1997, Pages 1497-1501 Copyright, International Associ~tion of Milk, Food and Environmental Sanitarians Effects of Temperature, ph, Glucose, and Citric Acid on the Inactivation of Salmonella typhimurium in Reduced Calorie Mayonnaise JEANNE-MARIE MEMBRE,l* VIRGINIE MAJCHRZAK? and ISABELLE JOLLy2 'L.G.P.T.A., 1nstitut National de la Recherche Agronomique, 369 rue Jules Guesde, BP 39,59651 Villeneuve d'ascq Cedex; and. 2SERMHA. 1nstitut Pasteur de Lille, 369 rue Jules Guesde, BP 39, 59651 Villeneuve d'ascq Cedex, France (MS# 96-307: Received 4 November 1996/Accepted 23 February 1997) ABSTRACT A Salmonella typhimurium strain was inoculated in reduced calorie mayonnaise. A central composite design was implemented to assess the effects of temperature (15 to 35 C), ph (4.5 to 6.5), glucose (1 to 4% [wt/vol]), and citric acid (0.05 to 0.1 % [wt/vol]) on the inactivation of Salmonella. Whatever the conditions, an inhibition of the strain was obtained, but only after a long period of time, from 11 to 85 days. In this study, as the survival curves obtained did not follow typical first -order destruction kinetics, the primary model chosen was exponential. A second-order polynomial linear regression was then used to study the effects of the various factors on the inhibition of S. typhimurium. Estimated values of the k parameter, which represented the shape of the destruction curves, were well correlated with the predicted ones (R2 = 0.94). Generally, the higher the temperature with a low ph, the greater the inactivation. With a citric acid concentration of 0.09% (wt/vol), no glucose effect could be seen. In contrast, a lower acid concentration, 0.06% (wt/vol), enabled the higher level of destruction to be reached with a 3.5% (wt/vol) glucose concentration. This study documented that reduced calorie mayonnaise containing citric acid can represent a nonnegligible consumer safety risk and indicated that a well-fitted model is of interest to correctly predict this risk. Key words: Microbial inactivation, modeling, Salmonella Generally, the predictive microbiological approach uses mathematical equations to estimate growth parameters such as the growth rate and lag period as affected by storage conditions (2, 24). Among the various models proposed (1, 7, 23, 28), the response surface technique is widely employed in food microbiology (5, 11, 26). For growth curves, this method leads to the establishment of a model which describes growth rate (log 11 or ~Il) as a function of temperature, ph, and N acl concentration (19). Few studies have reported modeling of microbial decline at suboptimal conditions. Recently, the survival of * Author for correspondence. Tel: 33 (0)3 2043 54 24; Fax: 33 (0)3 2043 5426; E-mail: membre@lille.inra.fr. Yersinia enterocolitica at low temperature was described by the vitalistic model (16). This latter equation, although empirical, allowed a good bacterial kinetic fit to be achieved and has been applied to thermal destruction kinetics of microorganisms like Listeria monocytogenes (6, 25) or Salmonella typhimurium (8). When several variables act in combination, linear or nonlinear models with three or four parameters have been proposed to fit microbial inactivation curves in a synthetic medium (3, 14, 15). In each case the parameters estimated with the primary model (2) were analyzed as functions of independent variables using the response surface technique. Few works have dealt with modeling nonthermal destruction directly in food products. Growth and survival of foodborne pathogens such as Yersinia enterocolitica, Salmonella spp., and Bacillus cereus were studied in cheese (17), but a great discrepancy between experimental data and data-based predictions was observed. In mayonnaise products the incidence of constituents on microbial contamination is quite well known: addition of an organic acid, particularly acetic acid, allows a good hygienic stability to be obtained when the acid produces a low ph value of around 3.5 (20). For instance, in commercial mayonnaise products containing acetic acid and with a ph value of less than 3.9, inactivation of Salmonella spp. was obtained whatever the mayonnaise formulation and temperature, even with an initial bacterial load of up to 10 6 CFU/ml (9, 18). Likewise, home-style salads prepared with low-ph mayonnaise and held at refrigeration temperature present minimal health hazard risks (10). In contrast, for formulations containing citric acid as organic acid the safety of mayonnaise has not been clearly established (20), and for mayonnaise in the ph range of 4 to 5 the antimicrobial properties of the ingredients and any synergistic effects that may occur have not yet been established (18). On the other hand, with a reduced calorie mayonnaise, to obtain an organoleptically acceptable product, a formulation change is necessary, mainly by increasing the ph value (20), The inclusion of citric acid in mayonnaise products could offer a favorable alternative in the preparation of reduced calorie products,

1498 MEMBRE, MAJCHRZAK, AND JOLLY Among the pathogenic microorganisms involved in mayonnaise spoilage, Salmonella enteritidis and Salmonella typhimurium are often reported (13, 14). The aim of this paper is to establish the response of Salmonella typhimurium to temperature, ph, glucose, and citric acid in reduced calorie mayonnaise and to use these data to predict how changes in formulation and storage conditions are likely to affect the survival of S. typhimurium in foods. MATERIALS AND METHODS Organism and culture conditions Salmonella typhimurium isolated from egg products was used throughout the study. The strain was subcultured on Trypticase soy agar (TSA) for 18 h at 37 C. For assays, commercial mayonnaise was used. Extra glucose and citric acid were added and the ph adjusted with 1 M NaOH as required by the experimental design. Each sample was inoculated to achieve an initial level of approximately 10 6 CFU/ml and incubated at the appropriate temperature for up to two months. TABLE 2. Explanatory variables and regression coefficients of the quadratic model for inactivation of Salmonella in the mayonnaise derivative model Factors Intercept Temperature ph Glucose Citric acid Temperature squared ph squared Citric acid squared ph X glucose ph X citric acid Glucose X citric acid 25 5.5 2.5 0.Q75 642 30.4 Variables" Regression coefficients Estimate 0.003 4.08-0.0013 0.408-0.0095 0.612 0.00006 0.0102-0.0016 205 0.0025 4.5 0.009 0.0015 0.0016-0.00013-0.00016-0.00011 Standard error 0.000076 0.0007 0.0015 0.0008 0.0007 0.001 0.0008 a The mean (11) and standard deviation (cr) of the levels of each explanatory variable were used in recording the factor levels as (value - Il)/cr. TABLE 1. Levels of the variables in the central composite design with four explanatory variables Temperature (0C) ph Glucose (%) Citric acid (%) 20 5.0 1.75 0.0625 20 5.0 1.75 0.0875 20 5.0 3.25 0.0625 20 5.0 3.25 0.0875 20 6.0 1.75 0.0625 20 6.0 1.75 0.0875 20 6.0 3.25 0.0625 20 6.0 3.25 0.0875 30 5.0 1.75 0.0625 30 5.0 1.75 0.0875 30 5.0 3.25 0.0625 30 5.0 3.25 0.0875 30 6.0 1.75 0.0625 30 6.0 1.75 0.0875 30 6.0 3.25 0.0625 30 6.0 3.25 0.0875 15 5.5 2.50 0.0750 35 5.5 2.50 0.0750 25 4.5 2.50 0.0750 25 6.5 2.50 0.0750 25 5.5 1.00 0.0750 25 5.5 4.00 0.0750 25 5.5 2.50 0.0500 25 5.5 2.50 0.1000 Experimental design A central composite design was implemented to assess the effects of temperature (15 to 35 C), ph (4.5 to 6.5), glucose (1 to 4% [wt/vol]) and citric acid (0.05 to 0.1 % [wt/vol]) on Salmonella typhimurium inactivation (Table I). To obtain orthogonal polynomials and thus a convenient interpretation of results, the factor levels were recorded as (value - Il)/cr where 11 is the mean parameter level and cr the standard deviation (Table 2). The data from each combination were the log of the number of bacteria for each sampling time interval. The model used to fit the survival curves was a model with only two parameters described as log N = (1 + log No) - exp (k t), where N represents the biomass quantity at time t and No the initial biomass quantity. The parameter k corresponds to the shape of the destruction curve: the higher the value of k, the greater the inactivation rate. The response variable is not N but log N. It was assumed that with a decimal logarithmic transformation the residual error would have a normal distribution. This model was preferred to the vitalistic one because it requires fewer parameters (two instead offour, respectively). To study the effects of temperature, ph, glucose, and citric acid on the inactivation curve, a second-order polynomial linear regression was used. Statistical analyses The nonlinear regression was computed with Splus software (AT&T Bell Laboratories, Murray Hill, NJ, USA), the parameters of the regression were estimated by the maximum likelihood method. The Gauss-Marquardt algorithm was used in the numerical approximation process (12). The linear regression was performed with SAS software (SAS Institute Inc., Cary, NC, USA). RESULTS AND DISCUSSION Survival curve fitting First, data from each of the 36 treatment conditions of the central composite design were analyzed separately. The curves were not typical first-order destruction kinetics: they were convex curves with an increasingly negative slope (1)

MODELING SALMONELLA SP. INACTIVATION 1499 a 7 7 6 Q6 5 =5,-, '8 -- -- 4 ~4 ~ ~ ~ 3 U 83 '-' 2 Z2 Z 1.$1.$ 0 0 b -1 400 800 1200 1600 0 400 800 1200 1600 7 Q6 =5 -- ~4 ~ 83 Z2.$1 c d 0 0 400 800 1200 800 I 1200 I 1600 FIGURE 1. Inactivation kinetics of Salmonella typhimurium in mayonnaise derivative model. Observed data (symbols) and predicted values obtained with the individual model to which the time curves were fitted one by one (dotted line) and with the complete model taking into account temperature, ph, glucose, and citric acid incidence (solid line),for (a) 20 o e, ph 5, 1.75% glucose, 0.0875% citric acid, (b) 20 o e, ph 6,3.25% glucose, 0.0625% citric acid, (c) 30 o e, ph 6,3.25% glucose, 0.0875% citric acid, and (d) 25 e, ph 5.5,4% glucose, 0.075% citric acid. (Fig. 1). Similar observations have been reported for Listeria monocytogenes inactivation in heated infant formula (15) or in an acid environment (4). The simple exponential equation (equation [1]) was effective in describing the biomass profile. In the literature, the vitalistic model (16) or the nonlinear model proposed by Whiting and Buchanan (27) have generally been chosen to describe death curves, but in our ranges of experimental conditions an equation containing four parameters was not necessary. Moreover, the model with two parameters attempts to respect properties listed by Ratkowsky (22): a simple model with as few parameters as possible, a parameterization providing stable parameters, a range of applicability extending over the entire experimental domain, a minimum variance of parameters, and an unbiased model with respect to statistical criteria. The parameter k can be regarded as an indicator of the rate of bacterial destruction. To assess the efficiency of treatment from inactivation curves, the time t 4D to achieve a bacterial population reduction of 4 log cycles, i.e., inactivation of 99.99% (3, 4), was calculated. As equation [1] does not correspond to first-order inactivation kinetics, t 4D was written as follows: t 4D = In 5/k. (2) Second, the parameter k, which can be regarded as an indicator of the rate of bacterial destruction, was determined as a function of temperature, ph, glucose, and citric acid concentrations. With a surface model method including linear, quadratic, and interactive effects of each factor, the correlation between k values estimated for each curve and k values predicted by the complete model was R2 = 0.94 (Fig. 2). One plotted point with a large residual corresponded to on g 0.004 '".", >.!l (,) 'i3 15.. 0.002..><: R 2 = 0.94 0.000 JI-------+------+----+--- 0.000 0.002 k calculated 0.004 values FIGURE 2. Predicted versus calculatedk parameter values for inactivation curves of Salmonella typhimurium in the mayonnaise model.

1500 MEMBRE, MAJCHRZAK, AND JOLLY the following combination: temperature 30 D C, ph 5, glucose 1.75%, and citric acid 0.0625%. This treatment led to bacterial inactivation in 500 h, but it was evaluated as 400 h. Though there was a substantial discrepancy between the calculated and predicted k values for this set of experimental conditions, this combination was retained in the statistical analysis. Only the significant terms (chi-square test, P < 0.05) were retained in the final equation. The values of the parameters and their standard errors are reported in Table 2. The experimental data points and the fitted values for the individual model (dotted line) and complete model (solid line) are plotted for various experimental conditions in Figure 1. Factor combination effect Predictive microbiology applied to food products can be used to describe the effects of environmental factors and their interactive effects on the survival of microorganisms. For reduced calorie mayonnaise, factors leading to inactivation are mainly low ph and high temperature and, to a lesser extent, high concentrations of glucose and citric acid. In fact, in the optimal region, the synergistic action of the ingredients inactivated S. typhimurium in three weeks whereas at 20 D C, ph 6, 1.5% glucose, and 0.09% citric acid the same bacterial destruction took two months. Moreover, the glucose effect depended on the citric acid concentration, i.e., the addition of glucose increased the bacterial destruction only at low citric acid concentrations. Mathematical equations have often been employed in the study of thermal destruction of microorganisms. The parameter commonly used is the decimal reduction time, D. To illustrate the inactivation effectiveness of the various environmental factors, t4d values are reported in Table 3. The use of orthogonal polynomials in the statistical analysis allowed a comparison of the relative importance of the linear, quadratic, and interactive terms. In this study the significant interactive effects, ph X glucose, ph X citric acid, and glucose X citric acid, were found to be quite similar (Table 2). A strong interactive effect between preservative agents has already been observed with a strain of Salmonella in a synthetic medium, in which a stimulatory effect of NaCI with acetic acid was noticed (21). Finally, temperature had a large positive effect on mayonnaise stability in the entire experimental domain. This can be explained as indicating that the higher the temperature, the more inhibitory the citric acid. Recently the log-logistic equation has been introduced to develop predictive approaches (6, 8), and in parallel there have been studies that reported modeling of microbial decline for combinations of suboptimal conditions (3, 4, 16). The survival of Yersinia enterocolitica was analyzed at low temperatures and growth inhibitory ph values in synthetic medium. Similarly, Listeria monocytogenes was studied in a reduced oxygen environment under varying conditions of temperature and lactic acid, sodium chloride, and sodium nitrite concentrations. The prediction of the bacterial inactivation of Salmonella typhimurium as a function of four environmental factors provides an additional opportunity for the use of the predictive microbiological approach. Food TABLE 3. Predicted mean values oft 4D for various levels of citric acid, glucose, ph, and temperature Citric Glucose Temperature t4d acid (%) (%) ph (DC) (days) 0.06 1.5 4.8 18 33.4 0.06 1.5 4.8 24 21.1 0.06 1.5 4.8 30 12.8 0.06 1.5 6.0 18 57.7 0.06 1.5 6.0 24 28.7 0.06 1.5 6.0 30 15.2 0.09 1.5 4.8 18 20.0 0.09 1.5 4.8 24 14.8 0.09 1.5 4.8 30 10.2 0.09 1.5 6.0 18 60.4 0.09 1.5 6.0 24 29.4 0.09 1.5 6.0 30 15.4 0.06 3.5 4.8 18 19.4 0.06 3.5 4.8 24 14.5 0.06 3.5 4.8 30 10.0 0.06 3.5 6.0 18 48.2 0.06 3.5 6.0 24 26.2 0.06 3.5 6.0 30 14.5 0.09 3.5 4.8 18 18.1 0.09 3.5 4.8 24 13.8 0.09 3.5 4.8 30 9.7 0.09 3.5 6.0 18 273.3 0.09 3.5 6.0 24 47.3 0.09 3.5 6.0 30 19.3 models have often been established in laboratory media. In such cases, challenge tests are necessary to assess the potential for use of the models as decision tools in the food industry. Studies directly carried out in food products have the advantage of producing an accurate model useful in food manufacture. The experimental design was selected to obtain results for mayonnaise products in a relatively high ph range. From an industrial point of view, this approach seems promising for evaluation of the risk of Salmonella spoilage in new products, such as in reduced calorie mayonnaise containing citric acid and glucose, and for development of hurdle technology directed against such spoilage. We demonstrated that bacterial contamination represents a nonnegligible consumer safety risk in reduced calorie mayonnaise. Even if the microbial activity is inhibited in each case, this inactivation can take 11 days and can then have a duration of two months. REFERENCES 1. Baranyi, J., T. A. Roberts, and P. McClure. 1993. A non-autonomous differential equation to model bacterial growth, Food Microbiol. 10:43-59. 2. Buchanan, R. L. 1993. Predictive food microbiology. Trends Food Sci. Technol. 4:6-11. 3. Buchanan, R. L., and M. H. Golden. 1995. Model for the non-thermal inactivation of Listeria monocytogenes in a reduced oxygen environment. Food Microbiol. 12:203-212. 4. Buchanan, R. L., M. H. Golden, and R. C. Whiting. 1993. Differentiation of the effects of ph and lactic or acetic acid concentration on the kinetics of Listeria monocytogenes inactivation. J. Food Prot. 56:474-478.

MODELING SALMONELLA SP. INACTIVATION 1501 5. Buchanan, R. L., and J. G. Phillips. 1990. Response surface model for predicting the effects of temperature ph, sodium chloride content, sodium nitrite concentration and atmosphere on the growth of Listeria monocytogenes. J. Food Prot. 53:370-376. 6. Cole, M. B., K. W. Davies, G. Mumo, C. D. Holyoak, and D. C. Kilsby. 1993. A vitalistic model to describe the thermal inactivation of Listeria monocytogenes. J. Ind. Microbiol. 12:232-239. 7. Dengremont, E., and J. M. Membre. 1995. Statistical approach for comparison of the growth rates of five strains of Staphylococcus aureus. Appl. Environ. Microbiol. 61:4389-4395. 8. Ellison, A., W. Anderson, M. B. Cole, and G. S. A. B. Stewart. 1994. Modelling the thermal inactivation of Salmonella typhimurium using bioluminescence data. Int. J. Food Microbiol. 23:467-477. 9. Erickson, J. P., and P. Jenkins. 1991. Comparative Salmonella spp. and Listeria monocytogenes inactivation rates in four commercial mayonnaise products. J. Food Prot. 54:913-916. 10. Erickson,1. P., D. N. McKenna, M. A. Woodruff, and J. S. Bloom. 1993. Fate of Salmonella spp., Listeria monocytogenes, and indigeneous spoilage microorganisms in home-style salads prepared with commercial real mayonnaise or reduced calorie mayonnaise dressing. J. Food Prot. 56:1015-1021. II. Gibson, A. M., N. Bratchell, and T. A. Roberts. 1988. Predicting microbial growth: growth responses of salmonellae in a laboratory medium as affected by ph, sodium chloride and storage temperature. Int. J. Food Microbiol. 6: 155-178. 12. Huet, S., E. Jolivet, and A. Messean. 1992. La regression non-lineaire, methodes et applications en biologie. INRA, Paris. 13. Humphrey, T. J. 1994. Contamination of egg shell and contents with Salmonella enteritidis: a review. Int. J. Food Microbiol. 21:31-40. 14. Leclair, K., H. Heggart, M. Oggel, F. M. Bartlet, and R. C. McKellar. 1994. Modelling the inactivation of Listeria monocytogenes and Salmonella typhimurium in simulated egg wash water. Food MicrobioI. 11:345-353. 15. Linton, R. H., W. H. Carter, M. D. Pierson, C. R. Hackney, and J. D. Eifert. 1996. Use of modified Gompertz equation to predict the effects of temperature, ph, and NaCI on the inactivation of Listeria monocytogenes Scott A heated in infant formula. J. Food Prot. 59:16-23. 16. Little, C. L., M. R. Adams, W. A. Anderson, and M. B. Cole. 1994. Application of a log-logistic model to describe the survival of Yersinia enterocolitica at sub-optimal ph and temperature. Int. J. Food Microbiol. 22:63-71. 17. Little, C. L., and S. Knochel. 1994. Growth and survival of Yersinia enterocolitica, Salmonella and Bacillus cereus in Brie stored at 4, 8 and 20 D C. Int. J. Food Microbiol. 24:137-145. 18. Lock, J. L., and R. G. Board. 1994. The fate of Salmonella entiritidis PT4 in deliberately infected commercial mayonnaise. Food MicrobioI. 11:499-504. 19. Membre, J. M., and P. M. Burlot. 1994. Effects of temperature, ph and NaCI on growth and pectinolytic activity of Pseudomonas marginalis. Appl. Environ. Microbiol. 60:2017-2022. 20. Radford, S. A., and R. G. Board. 1993. Review: fate of pathogens in home-made mayonnaise and related products. Food Microbiol. 10:269-278. 21. Radford, S. A., and R. G. Board. 1995. The influence of sodium chloride and ph on the growth of Salmonella entiritidis PT4. Lett. Appl. Microbiol. 20:11-13. 22. Ratkowsky, D. A. 1993. Principles of nonlinear regression modeling. J. Ind. Microbiol. 12:195-199. 23. Skinner, G. E., and J. W. Larkin. 1994. Mathematical modeling of microbial growth: a review. J. Food Safety 14:175-217. 24. Sofos, J. N. 1993. Current microbial considerations in food preservation. Int. J. Food Microbiol. 19:87-108. 25. Stephens, P. J., M. B. Cole, andm. V. Jones. 1994. Effect of heating on the thermal inactivation of Listeria monocytogenes. J. Appl. Bacteriol. 77:702-708. 26. Thayer, D. w., W. S. Muller, R. L. Buchanan, and J. G. Phillips. 1987. Effect of NaCI, ph, temperature and atmosphere on growth of Salmonella typhimurium in glucose mineral salts medium. Appl. Environ. Microbiol. 53:1311-1315. 27. Whiting, R. C.,andR. L. Buchanan. 1992. Use of predictive microbial modeling in a HACCP program, p. 125-141. In Proceedings of the Second ASEPT International Conference on Predictive Microbiology andhaccp. 28. Zwietering, M. H., T. Wijtzes, J. C. De Wit, and K. Van't Riet. 1992. A decision support system for prediction of the microbial spoilage in foods. J. Food Prot. 55:973-979.