Predictive Modeling of Psyctirotrophic Bacillus cereus

Size: px

Start display at page:

Download "Predictive Modeling of Psyctirotrophic Bacillus cereus"

Damian Mosley
6 years ago
Views:

1 684 Journal of Food Protection, Vol. 56, No. 8, Pages (August 1993) Copyright, International Association of Milk, Food and Environmental Sanitarians Predictive Modeling of Psyctirotrophic Bacillus cereus JANICE M. BAKER and MANSEL W. GRIFFITHS* Department of Food Science, University of Guelph, Guelph, Ontario, Canada NIG 2WJ (Received for publication October 7, 1992) ABSTRACT Response surface analysis was used to determine the effects and interactions of water activity (0.965 to 0.995), ph (5.8 to 8.0), temperature (6 to 38 C), glucose concentration (0 to 1.8%), and starch concentration (0 to 0.625%) on the growth of and toxin production by a psychrotrophic strain of Bacillus cereus in brain heart infusion broth. Growth was measured by monitoring optical density and plate counts, toxin production was assayed by an immunological method (BCET-RPLA toxin assay) and cytotoxicity with Vero and HEp-2 cells. Regressions were performed using response surface techniques, on Growth, LnGrowth, RPLA, LnRPLA, Vero, LnVero, and HEp-2; quadratic predictive equations for growth and toxin production were obtained. The results indicate the factors that had the greatest influence on both growth and toxin production were water activity and temperature. Predicted values obtained from the model were in good agreement with experimental values. Bacillus cereus is a gram-positive, aerobic, rod-shaped organism, that can produce spores capable of surviving pasteurization temperatures. Indeed, it has been found that pasteurization can lead to activation of spores. On germination, vegetative cells of psychrotrophic strains can grow at refrigeration temperatures (12). B. cereus can produce two distinct toxins; a heat-labile diarrhegenic toxin and a heat-stable emetic toxin (10). The organism has been implicated in foodborne illness, with rice and other starchy foods, dairy products, and meat and vegetable products being the most common vectors (10,16,17). In Canada, between the years 1975 and 1986, the number of reported cases of foodborne illness caused by B. cereus increased from 1 to 106. However, the number of outbreaks in the same time period remained low (16,17). Psychrotrophic strains of B. cereus are common contaminants of raw and pasteurized milk (5,9,12,18). Of 24 commercial pasteurized milk cartons tested in our laboratory, 12 cartons produced 22 isolates of B. cereus (Baker and Griffiths, unpublished results). Seventeen of the 22 isolates were capable of producing toxin in brain heart infusion (BHI) broth after 7 d at 6 C making these organisms a concern to food processors. Acquiring some knowledge of the nutritional and environmental factors that affect growth, and toxin production by B. cereus, may aid in developing methods for controlling this organism during food processing and storage. Previous investigations have focused on mesophilic strains of B. cereus and have monitored the effect of environmental and nutrient conditions singly, and in isolation (7). These studies gave no consideration to any interactions that may occur between environmental or nutritional factors, as is the case in a food system or processing plant environment. In food microbiology, predictive modeling refers to systematically measuring growth responses of microorganisms and expressing these responses via one of a number of mathematical models. Once the model is validated in the types of products in which it will be used, it can be used repeatedly to accurately predict the growth of the microorganism under a number of conditions, within the limits of the original experimentation (4). Predictive modeling has been shown to be useful in defining microbiologically safe operating practices, such as conditions for a critical control point in a hazard analysis critical control (HACCP) program, or predicting the growth of a microorganism in a new formulation of a product (4). Predictive models have been fully developed for organisms such as Salmonella (8), Shigella flexneri (19), Listeria monocytogenes (2), and Aeromonas hydrophila (13) and incorporated into a user friendly computer program. Bacillus species have not been investigated as much as other microorganisms in terms of developing a predictive model, perhaps because they are not seen as being as hazardous as the previously mentioned bacteria. However, B. cereus is an important spoilage and pathogenic organism, and a predictive model would prove useful to the food processing industry. The objective of this study was to determine which factors or interaction of factors were significant to the growth of and toxin production by a psychrotrophic strain of B. cereus, and to incorporate this information into the formulation of a predictive model for each phenomenon. MATERIALS AND METHODS Strain and inoculum preparation The strain of B. cereus used was isolated from an Ontario raw milk sample obtained from the Central Milk Testing Lab in Guelph, Ontario. It was grown for 18 h in BHI broth (Difco Laboratories, Detroit, MI) at 30 C and was inoculated into the experimental broths at a level of 0.1% for a cell density of ca. 1 x 10 4 CFU/ml. Gram and spore stains were performed on the inocula to ensure that it consisted of vegetative cells. BHI broth was used as a basal medium throughout the experiment.

2 PREDICTIVE MODELING OF B. CEREUS 685 Experimental design Data were obtained using a central composite rotatable response surface design. The 2 5 factorial layout was run in six incomplete blocks to reduce the week-to-week variability of the experiment. The effects of water activity (0.965, 0.973, 0.980, 0.988, 0.995), ph (5.8, 6.4, 6.9, 7.5, 8.0), temperature (6, 14, 22, 30, 38 C), glucose concentration (0, 0.45, 0.9, 1.35, 1.8%), and starch concentration (0, 0.156, 0.315, 0.469, 0.625%) were investigated simultaneously. Replicates were built into the design. Growth media alteration Sodium chloride (5.4, 4.04, 2.84, 1.48, or 0 g) was added to BHI powder (3.7 g) to obtain the water activities 0.965, 0.973, 0.980, 0.988, and 0.995, respectively. The ph was adjusted to 5.8, 6.4, 6.9, 7.5, and 8.0, using a phosphate buffer system consisting of a 0.2 M solution of monobasic potassium phosphate and a 0.2 M solution of dibasic sodium phosphate. The glucose was added as D- glucose. The starch (0, 6.25, 12.5, 18.75, or 25 ml of a 2.5% soluble starch solution) was added to obtain 0, 0.156, 0.315, 0.469, or 0.625% starch, respectively. The broths were made up to 100 ml with distilled water in glass bottles and sterilized by autoclaving for 15 min at 121 C. The broths were inoculated as described above and incubated at the appropriate temperature in incubators or water baths without agitation. The water activities were confirmed after autoclaving by freezing point depression, using a cryoscope (Advanced Instruments Inc., Newton Highlands, MA) according to the method of Ferro Fontan and Chirife (6). Growth monitoring Samples of broth were aseptically removed and absorbance at 600 nm determined using a spectrophotometer (Shimadzu, Kyoto, Japan). Plate counts were performed on milk agar (Oxoid, Nepean, Ontario) using a Spiral Plate Maker (Spiral System Instruments, Inc., Bethesda, MD). Duplicate plates were counted after 24 h of incubation at 22 C. Growth was measured as the time required for the optical density to reach a level of 0.2 units, which related to approximately 3 x 10 6 cells per ml. Growth was measured for 477 h. Assay of toxin production Diarrheal toxin production was measured by the BCET Reverse Passive Latex Agglutination (RPLA) kit (Oxoid) following the instructions provided by the manufacturer and by a cytotoxicity method using Vero (African Green Monkey Kidney) cells according to the method of MacLeod and Gyles {11). This method was also used to measure emetic toxin production using HEp-2 cells (Agriculture Canada, Guelph, Ontario), but in this case, the samples were subjected to a heat treatment of 121 C for 20 min (15) before the cell culture assay was performed. Statistical analysis Data analysis was performed using least squares analysis of the general linear model procedure of the SAS system (14). The regression analysis was performed on the untransformed data of Growth, RPLA, Vero, and HEp-2; on natural logarithm transformations of the Growth, RPLA, and Vero data; and on the reduced LnGrowth, LnRPLA, and LnVero equations. Only one sample did not exhibit growth by 477 h; thus, it was treated as a missing value in the analysis. RESULTS AND DISCUSSION Predictive equations Growth and toxin production data were used to obtain predictive equations for a psychrotrophic strain of B. cereus. The effect of the variables temperature, ph, water activity, starch, and glucose on the growth response and toxin producing capabilities of B. cereus was assessed. The growth and toxin data were analyzed using response surface methodology to obtain quadratic predictive equations for growth, and toxin production as assayed by the RPLA toxin assay kit, the Vero cell cytotoxicity assay, and the HEp-2 cell cytotoxicity assay. The R 2 values for these models, which indicate how well the derived models fit the data used to generate them, are shown in Table 1. The low R 2 value for the model predicting the production of emetic toxin as assayed by the HEp-2 cytoxicity assay may be due to the fact that the organism used did not produce emetic toxin. Although the toxin detected in this assay was most certainly heat stable, the data gathered were not adequate for the generation of a predictive model for emetic toxin production, and thus, it will not be discussed further. A natural logarithm transformation of the growth and toxin data was performed in order to make the data fit the normal distribution. The resulting models are shown in Table 2. The R 2 values increased for LnGrowth and LnVero and TABLE 1. Comparison ofr 2 values for the untransformed models. Model R 2 Growth RPLA Vero HEp TABLE 2. Quadratic response surface models for the effects and interactions of temperature (temp) ( C), ph, water activity (a J, starch concentration (%), and glucose concentration (glu) (%) on the (a) Growth of, and (b & c) Diarrhegenic toxin production by a psychrotrophic strain of B. cereus. a) LnGrowth = temp pH a w starch glu temp pH a w starch glu (temp * ph) (temp * a w ) (temp * starch) (temp * glu) (pH * a w ) (pH * starch) (pH * glu) (a w * starch) (a w * glu) (starch * glu) R 2 = n = 61 F value = Pr>F = b) Vero cell toxicity assay LnVero = temp pH a w starch glu temp pH a w starch glu (temp * ph) (temp * aj (temp * starch) (pH * a w ) (pH * starch) (pH * glu) (a w * starch) (a w * glu) R 2 = n = 62 F value = 8.30 Pr>F = c) Reverse Passive Latex Agglutination assay LnRPLA = temp pH a w starch glu temp pH a w starch glu (temp * ph) (temp * a w ) (temp * starch) (temp * glu) (pH * aj (pH * starch) (pH * glu) (a w * starch) (a w * glu) (starch * glu) R 2 = n = 62 F value = 3.50 Pr>F = JOURNAL OF FOOD PROTECTION. VOL. 56, AUGUST 1993

3 686 BAKER AND GRIFFITHS decreased slightly for LnRPLA, indicating that better fitting models for growth and the Vero cell toxin assay were obtained after transformation. Transformation did not significantly alter the model for the RPLA assay. Factors other than the R 2 value come into play when deciding which model to use for the pediction of toxin production based on the RPLA assay. The skewness and kurtosis decrease and the distribution of the data points are more normal after the Ln transformation, thus suggesting that the LnRPLA model is better than the untransformed model. In order to simplify the models, abbreviated models consisting of all the single factors, all the squared factors and those interactions of factors that were significant at P < 0.15 were generated (Table 3). When using experimental growth data from broths composed of variable combinations not used TABLE 3. Reduced quadratic response surface models for the effects and interactions of temperature (temp) ( C), ph, water activity (a J, starch concentration (%), and glucose concentration (glu) (%) on the (a) Growth of and (b & c) Diarrhegenic toxin production by a psychrotrophic strain of B. cereus. a) LnGrowth = temp pH a w starch glu temp pH aJ + starch glu (temp * ph) (temp * a ) * R 2 = Pr>F = b) Vero cell toxicity assay LnVero = temp pH a w lstarch glu temp pH a w starch glu 2 R 2 = n = 62 F = Pr>F = c) Reverse Passive Latex Agglutination assay LnRPLA = temp pH a w starch glu temp pH aJ starch glu (temp * a w ) R 2 = n = 62 F = 5.82 Pr>F = = to generate the model, the abbreviated LnGrowth model resulted in the best correlation with the experimental data (R 2 = 0.988). The F and P values for the abbreviated models for LnGrowth, LnVero, and LnRPLA are shown in Table 4. The factors playing a major role in growth and toxin production are temperature and water activity. The two most significant factors for all of the models are composed of either the single term or squared term or the interaction of temperature and water activity. Glucose, ph, and starch are significant to a lesser degree in some of the models. Other than the interaction between temperature and water activity, no interaction terms were found to significantly affect growth or toxin production. It may be that the levels of the variables tested in this study were not in a range suitable for observation of interactive effects. The single and squared terms of ph, water activity, and starch were found to be insignificant in the LnRPLA model. These factors were included in the model in order to determine optimum values for each of the factors and to make for a more practical model. Different ranges of these nonsignificant factors should be tested to see if that would increase their significance for the RPLA assay. Also, questions have been raised (3,18) about the specificity of the BCET-RPLA assay. This lack of specificity for the diarrhegenic toxin of B. cereus may have contributed to some false-positive or false-negative results in the data collection, thus affecting the model. The optima for each factor of the models are listed in Table 5. Other than the ph value for the LnRPLA model being outside of the range tested (which may be due to reasons mentioned above), the optima for the two toxin assay models are similar. Optimal growth and toxin production occurred at equivalent glucose concentrations, but in all other cases, the optimum for growth was higher than that required for maximal toxin production. Response surface graphs (Fig. la and lb) demonstrate the effect of starch on growth. Fig. la shows the response of the LnGrowth abbreviated model at ph 7 in the absence of glucose and starch. In the presence of 0.55% starch (Fig lb), TABLE 4. s for variables 'for abbreviated LnGrowth, L nvero, and LnRPLA models, based on the Type III sum s of squares. Growth Diarrhegenic toxin (Vero cell assay) Diarrhegenic toxin (RPLA ass ay) Variable Pr>P Pr>F Pr>F Temp 8.69 ph 4.44 a w 5.76 Starch 1.56 Glu Temp ph a w 5.61 Starch Glu Temp*a w Temp*pH * * * * * * * * * * * * * * * * * * * * Abbreviations as in Table 2. Probability of a larger value of F. Those variables significant at P < 0.05 are indicated by an asterisk (*).

PREDICTIVE MODELING OF B. CEREUS 687 TABLE 5. Optimum conditions for growth and toxin production based on abbreviated models LnGrowth, LnVero, and LnRPLA.

4 PREDICTIVE MODELING OF B. CEREUS 687 TABLE 5. Optimum conditions for growth and toxin production based on abbreviated models LnGrowth, LnVero, and LnRPLA. Variable Temperature ph Water activity Starch Glucose LnGrowth ** Outside range of values tested. r- n I LnVero LnRPLA 24.5 ** Figure la. Response surface graph showing the effects of temperature and water activity on growth at ph = 7, 0% starch, 0% glucose. cereus was near optimum at 1% starch, with a temperature of 30 C, and water activity of BHI broth (0.995). As stated earlier, studies performed previously on factors affecting the growth of B. cereus have examined these singly and in isolation (7). The data from such an approach are of limited value. By studying many factors at once, results more realistically reflect conditions that apply in foods where there is an interaction of a number of conditions on the microorganisms present. Developing a model from data obtained in this way will provide the microbiologist or food processor with a valuable tool that will allow the accurate prediction of events actually occurring in a food product or a processing environment. The results of this study, one of the first to model the effects of more than one factor at a time on the growth of and toxin production by B. cereus, have shown that interactions of nutritional and environmental conditions are important when predicting growth. As in the development of any model, the next step is validation. Due to the importance of B. cereus in dairy products, further investigations will examine the validity of the above equations in dairy products. An immunological based assay has recently been developed, (TECRA Bacillus Diarrhoeal Enterotoxin Visual Immunoassay, Bioenterprises Pty Ltd., Roseville, New South Wales, Australia), and a model based on this assay is being prepared. More investigations still need to be done on developing rapid, definitive assays for diarrhegenic and emetic toxin production before a reliable predictive model can be developed for toxin production. Additional levels of the variables studied and comparisons in foods need to be carried out in order to further evaluate and refine the models. However, this study has provided valuable information for determining the probability of growth of B. cereus under a variety of conditions. ACKNOWLEDGMENTS v Q 1 Figure lb. Response surface graph showing the effects of temperature and water activity on growth at ph = 7, 0.55% starch, 0% glucose. the entire curve is shifted downward, indicating that an optical density of 0.2 is reached in a shorter time period. This stimulatory effect of starch supports the findings of Garcia- Arribas and Kramer (7), who showed that growth of B. This research is supported by the Ontario Milk Marketing Board. The authors wish to express their thanks to Dr. O. Brian Allen, Dept. of Mathematics and Statistics, University of Guelph, for his assistance with the experimental design and statistical analysis of the data. This paper was presented in part, at the 79th Annual IAMFES Meeting, Toronto, Ont., July 26-29, REFERENCES 1. Buchanan, R. L Using spreadsheet software for predictive microbiology applications. J. Food Safety 11: Buchanan, R. L., and J. G Phillips 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: , Buchanan, R. L., and F. J. Schultz Evaluation of the Oxoid BCET-RPLA kit for the detection of Bacillus cereus diarrheal enterotoxin as compared to cell culture cytotonicity. J. Food Prot. 55: Cole, M Predictive modelling-yes it is! Lett. Appl. Microbiol. 13: Christiansson, A., A. Satyanarayan Naidu, I. Nilsson, T. Wadstrom, and H-E. Pettersson Toxin production by Bacillus cereus dairy isolates in milk at low temperatures. Appl. Environ. Microbiol. 55: Ferro Fontan, C, and J. Chirife The evaluation of water activity in aqueous solutions from freezing point depression. J. Food Technol. 16: Garcia-Arribas, M. L., and J. M. Kramer The effect of glucose,

5 688 BAKER AND GRIFFITHS starch, and ph on growth, enterotoxin and haemolysin production by strains of Bacillus cereus associated with food poisoning and nongastrointestinal infection. Int. J. Food Microbiol. 11: Gibson, A. M., N. Bratchell, and T. A. Roberts Predicting microbial growth: responses of salmonellae in a laboratory medium as affected by ph, sodium chloride and storage temperature. Int. J. Food Microbiol. 6: Griffiths, M. W Toxin production by psychrotrophic Bacillus spp. present in milk. J. Food Prot. 53: Kramer, J. M., and R. J. Gilbert Bacillus cereus and other Bacillus species, pp In M. P. Doyle (ed.), Foodborne bacterial pathogens. Marcel Dekker, New York. 11. MacLeod, D. L., and C. L. Gyles Purification and characterization of an Escherichia coli shiga-like toxin II variant. Infect. Immun. 58: Meer, R. R., J. Baker, F. W. Bodyfelt, and M. W. Griffiths Psychrotrophic Bacillus spp. in fluid milk products: A review. J. Food Prot. 54: Palumbo, S. A., A. C. Williams, R. L. Buchanan, and J. G. Phillips Model for the aerobic growth of Aeromonas hydrophila K144. J. Food Prot. 5: SAS Institute, Inc SAS/STAT guide for persona] computers, Version 6 Ed. SAS Institute, Inc., Cary, NC. 15. Szabo, R. A., J. I. Spiers, and M. Akhtar Cell culture detection and conditions for production of a Bacillus cereus heat-stable toxin. J. Food Prot. 54: Todd, E. C. D Foodborne disease in Canada: A 10-year summary, Polyscience Publications, Inc., Ottawa. 17. Todd, E. C. D Foodborne and waterborne disease in Canada: Annual summaries, 1985 and Polyscience Publications, Inc., Ottawa. 18. Van Netten, P., A. Van De Moosdijk, P. Van Hoensel, and D. A. A. Mossel Psychrotrophic strains of Bacillus cereus producing enterotoxin. J. Appl. Bacteriol. 69: Zaika, L. L., J. G. Phillips, and R. L. Buchanan Model for aerobic growth of Shigella flexneri under various conditions of temperature, ph, sodium chloride, and sodium nitrite concentrations. J. Food Prot. 55:

Guidance on the safety and shelf-life of vacuum and modified atmosphere packed chilled foods. January 2004 (DRAFT)

Guidance on the safety and shelf-life of vacuum and modified atmosphere packed chilled foods January 2004 (DRAFT) Introduction This document provides advice on vacuum and modified atmosphere packaged (VP/MAP)