ABSTRACT

The objectives of this study were to develop a probability model of Staphylococcus aureus enterotoxin A (SEA) production as affected by water activity (aw), pH, and temperature in broth and assess its applicability for milk. The probability of SEA production was assessed in tryptic soy broth using 24 combinations of aw (0.86 to 0.99), pH (5.0 to 7.0), and storage temperature (10 to 30°C). The observed probabilities were fitted with a logistic regression to develop a probability model. The model had a concordant value of 97.5% and concordant index of 0.98, indicating that the model satisfactorily describes the probability of SEA production. The model showed that aw, pH, and temperature were significant factors affecting the probability of toxin production. The model predictions were in good agreement with the observed values obtained from milk. The model may help manufacturers in selecting product pH and aw and storage temperatures to prevent SEA production.

Staphylococcus aureus is a foodborne pathogen that frequently contaminates food products during preparation and processing (12). Human staphylococcal food poisoning (SFP) is caused by the ingestion of heat-stable staphylococcal enterotoxins (SE) produced by S. aureus (16). Symptoms of SFP include nausea, vomiting, and stomachache, which can occur within 2 to 8 h after toxin ingestion (1). Patients with this disease often need hospitalization, especially immunocompromised individuals.

Numerous cases of SFP have been reported. In Japan, there were 13,420 SFP cases caused by contaminated dairy products in 2000 (2). In China, a survey that examined the contamination levels of S. aureus in raw milk in Heilongjiang Province showed that the prevalence of S. aureus was 83.5% in 400 raw milk samples (27). According to data from the National Foodborne Diseases Surveillance Network, there were 94 outbreaks and 1,186 hospitalizations of 2,223 confirmed SFP cases from 2003 to 2007 in China (14). In 2008, 119 students in Shenzheng, China, a metropolis near the east coast of China, were sickened by SFP after drinking a contaminated milk product (10). More recently, an outbreak of SFP occurred in Deyang, China, with 70 individuals infected from consuming staphylococcal toxin-contaminated milk (28). Due to the typical quick recovery from SFP, unreported minor outbreaks, or misdiagnosis, the actual number of SFP cases is likely to be much higher (9). Since S. aureus can cause serious health issues, controlling this pathogen at all stages of the food chain is of public health, economic, and social importance.

Previous studies have indicated that SE would likely be produced when the population of S. aureus reaches >105 CFU/ml (11, 18, 24). However, there were other studies showing that S. aureus at >105 CFU/ml was not necessarily an indication of SE production. For example, Rajkovic et al. (17) found that SE was produced in sandwich components when the S. aureus counts were below 103 CFU/g. Several studies have indicated that inoculum size, temperature, pH, and water activity (aw) affect the SE production in food products (6, 7, 13, 19, 23). For example, Schelin et al. (20) reported that SE production occurred at pH 5.0 to 9.6, with the optimal pH being 7 to 8, and at aw above 0.86, with the optimal aw being 0.99.

Growth or no-growth models of S. aureus in food as affected by environmental factors have been developed (3, 25). To our knowledge, there is no published report regarding the probability of SE production as affected by aw, pH, and temperature. Therefore, the objectives of this study were to develop a probability model for SE toxin production as a function of aw, pH, and temperature and evaluate its applicability for ultra-high-temperature-processed milk.

MATERIALS AND METHODS

Preparation of S. aureus strain.

S. aureus strain CICC 10786 (S. aureus enterotoxin A [SEA] positive) from Shanghai Municipal Center for Disease Control and Prevention was used in this study. The culture was initially grown on Baird-Parker agar (Hangzhou Microbial Reagent Co., Hangzhou, China) at 37°C for 24 h. Cells of several well-grown colonies on the plate were transferred to 50 ml of tryptic soy broth (TSB; Qingdao Hope Bio-Technology Co., Qingdao, China) and incubated at 37°C for 24 h with shaking at 150 rpm to reach the stationary phase. The cell suspension was centrifuged at 5,000 × g for 10 min at 4°C and washed twice with sterile 0.85% saline solution. The pellet was diluted with 0.85% saline to approximately 5.0 log CFU/ml for use as inoculum.

Sample preparation.

Twenty-four treatments (Table 1) were examined to determine the probabilities of SEA production in TSB of various aw (0.86 to 0.99) and pH (5.0 to 7.0) values at storage temperatures of 10 to 30°C. The treatments were selected based on the reported growth boundary for S. aureus and SEA production (15, 23, 25). The pH of TSB was adjusted with 0.1 M HCl or 0.1 M NaOH, and the aw was adjusted with sodium chloride (Sinopharm Chemical Reagent Co., Shanghai, China) measured by an Aqualab 4TE aw meter (Decagon Devices, Pullman, WA). An S. aureus inoculum (0.5 ml) was added to 50 ml of sterile pH- and aw-adjusted TSB. The initial population of S. aureus in TSB was approximately 3.0 log CFU/ml, enumerated by plate count agar (Hangwei Co., Hangzhou, China). Samples were stored at the selected temperatures (Table 1) for 7 days. Two trials were conducted. In each trial, 10 samples were prepared for each treatment.

TABLE 1.

Treatments examined, probabilities of S. aureus enterotoxin production in TSB, observed and predicted probabilities for enterotoxin production, and final populations of S. aureus

Treatments examined, probabilities of S. aureus enterotoxin production in TSB, observed and predicted probabilities for enterotoxin production, and final populations of S. aureus
Treatments examined, probabilities of S. aureus enterotoxin production in TSB, observed and predicted probabilities for enterotoxin production, and final populations of S. aureus

Detection of SEA.

The presence of SEA in samples at the end of storage was determined by using a Ridascreen SET Total S. aureus enterotoxin immunoassay kit (R-Biopharm AG, Darmstadt, Germany) according to the manufacturer's instructions. The assay's detection limit for SEA is 0.25 ng/ml. A sample that was positive according to the cut-off absorbance value was recorded as a toxin-contaminated sample. The probability of toxin production for a treatment was the ratio of the number of toxin-contaminated samples to the total number of samples.

Model fitting.

The probabilities of SEA production for the 24 treatments were fitted with a linear logistic regression model, using the Logistic procedure of SAS 9.1.2 (SAS Institute, Cary, NC), as follows:

formula

where a0 to a9 are the coefficients to be estimated, pH is the broth pH, , which was transformed to stabilize the variance and provide a more suitable model (8), and T is the incubation temperature (°C). Since Logit(p) is equal to ln[p/(1 − p)], the probability of toxin production (p) can be estimated by the equation p = 1/[1 + e−Logit(p)].

Validation of model performance.

Model validation is an essential step in establishing a predictive model for practical uses. The validation includes internal and external validations. In internal validation, goodness-of-fit indices, including concordance, discordance, and the concordance index, c (26), were used to assess the agreement between the predicted values and observed values. The external validation step is commonly conducted using independent data for selected conditions within the experimental design range (aw of 0.86 to 0.99 and pH of 5.0 to 7.0 at 10 to 30°C) not included in model development (21). Therefore, the probability of SEA production in an additional 5 treatments in TSB and 3 treatments in milk (Table 2) was determined. The probability data used to validate the model employed indices that included median relative error (MRE) and mean absolute relative error (MARE) (4), which is determined as follows:

formula

where Xo is the observed value and Xp is the predicted value.

TABLE 2.

Comparison of predicted and observed probabilities in TSB broth and milk

Comparison of predicted and observed probabilities in TSB broth and milk
Comparison of predicted and observed probabilities in TSB broth and milk

RESULTS AND DISCUSSION

Effect of aw, pH, and temperature on S. aureus growth and SEA production.

The SEA production and final populations of S. aureus in 24 treatments are shown in Table 1. The results indicated that the populations of S. aureus in TSB ranged from 2.6 to 9.1 log CFU/ml after 7 days of incubation. When the final population of S. aureus was at approximately 3.0 log CFU/ml, no SEA was detected. When the population was greater than 5.2 log CFU/ml, SEA was detected. Based on the cell population and toxin production, it is reasonable to assume that when the population of S. aureus in TSB is greater than 5.0 log CFU/ml, SEA is likely to be produced. It is recognized that SE production is related to the growth of S. aureus. Notermans and Heuvelman (15) reported that SEA was produced in brain heart infusion broth under nearly all conditions of aw of 0.87 to 0.99, pH 4.0 to 7.0, and temperature of 8 to 30°C that allowed the growth of S. aureus, whereas Fujikawa and Morozumi (7) reported that SE was produced in milk after the S. aureus population reached 6.5 log CFU/ml. In addition, SE production is also influenced by the S. aureus strain and food matrix (17, 23).

Probability of toxin production and model development.

The average probability of SEA production for each treatment is listed in Table 1. Among the 24 treatments, 12 treatments had no toxin production (p = 0%), and 5 treatments had toxin production in 10 samples (p = 100%). The probabilities of toxin production were fitted with a logistic model, as follows:

formula

The coefficients and corresponding standard errors and p values are presented in Table 3. The concordance, discordance, and c values were 97.5%, 1.6%, and 0.98, respectively. The concordance value is the ratio of the number of pairs of observations that are concordant to the total number of pairs of observations, and c ranges from 0.5 to 1, with 1 indicating that the model perfectly predicts the responses. When the concordance is 100%, discordance is 0%, and c is 1, the predicted and observed values are completely in agreement (5). The results indicated that the probability model had a good predictive ability. The bw, bw × pH, bw × T, and pH × T and the quadratic term of bw significantly affect (P < 0.05) the probability of toxin production. Comparing the predicted and observed probabilities in Table 1, the predicted values are in agreement with the observed values for the majority of the treatments. The prediction overestimated the probability for the treatment with aw of 0.96 and pH of 6.8 at 18°C and underestimated the probabilities for treatments with aw of 0.93 and pH of 6.7 at 18°C and aw of 0.93 and pH of 5.2 at 25°C.

TABLE 3.

Parameter estimates of the logistic regression

Parameter estimates of the logistic regression
Parameter estimates of the logistic regression

Assuming that P ≤ 0.1 means a “no SEA production” region, P > 0.5 means an “SEA production” region, and 0.1 < P ≤ 0.5 means an “SEA likely to be produced” region, which are similar to the growth, likelihood of growth, and no-growth boundaries of foodborne pathogens (22, 25), the predicted SEA production and no-production boundaries at T of 20°C, pH of 6.0, or aw of 0.96 are shown in Figure 1. The SEA production and no-production predictions indicated that the conditions for SEA production were pH of >5.0 and aw of >0.86 at temperatures of >15°C. Similar results were reported by Schelin et al. (20), who found that the enterotoxin production in ultra-high-temperature-processed milk was restricted at a pH of <5.0 and aw of <0.86 at temperatures of ≤14°C.

FIGURE 1.

Boundaries for predicted “no SEA production” (P ≤ 0.1) and “SEA production” (P ≥ 0.5) at T of 20°C (A), pH of 6.0 (B), and aw of 0.96 (C). Probabilities obtained from treatments with the same T, pH, or aw (Table 1) were superimposed to compare the predicted and observed probabilities (⋄, no SEA production;likely SEA production; ♦, SEA production).

FIGURE 1.

Boundaries for predicted “no SEA production” (P ≤ 0.1) and “SEA production” (P ≥ 0.5) at T of 20°C (A), pH of 6.0 (B), and aw of 0.96 (C). Probabilities obtained from treatments with the same T, pH, or aw (Table 1) were superimposed to compare the predicted and observed probabilities (⋄, no SEA production;likely SEA production; ♦, SEA production).

Model validation.

In China, milk consumption is becoming popular. Milk serves as a suitable growth medium for microorganisms, and S. aureus is a pathogen of particular concern for milk safety. Cases of SFP are frequently linked to the consumption of milk and milk products (15). Therefore, a model validation experiment was conducted using TSB and ultra-high-temperature-processed milk with treatments not included in the model development (Table 2). The predictive ability of the model for validation treatments was assessed by MRE and MARE. The MRE of the model predictions was used as a measure of bias by estimating the mean difference between the observed and predicted values. MARE was used to measure the model's predictive accuracy, which assesses how close the predicted values are to the observed values. The results showed that the predicted probabilities were close to the measured values, and the predictions were acceptable (MRE = 0.030334 and MARE = 0.332909). Especially for milk, the model predictions were in agreement with the observed probabilities, indicating a good applicability of the probability model to milk.

In conclusion, the probability model for SEA production developed in this study provided reasonable predictions for the probabilities of SEA production at the ranges of aw, pH, and temperature tested and identified the aw, pH, and temperature limits for SEA production. The model may be used to select aw, pH, and storage temperatures for milk that prevent the production of SEA by S. aureus. Further studies are needed to evaluate the applicability of this model to other types of food products.

ACKNOWLEDGMENTS

This study was supported by the National Natural Science Funds of China (grant no. 31401608 and 31271896) and Chinese National Key Technology R & D Program (grant no. 2012BAD29B07).

REFERENCES

REFERENCES
1.
Ángeles
,
M. A.
,
M. M.
Carmen
, and
M. R.
Rosario
.
2010
.
Food poisoning and Staphylococcus aureus enterotoxins
.
Toxins
2
:
1751
1773
.
2.
Asao
,
T.
,
Y.
Kumeda
,
T.
Kawai
,
T.
Shibata
,
H.
Oda
,
K.
Haruki
,
H.
Nakazawa
, and
S.
Kozaki
.
2003
.
An extensive outbreak of staphylococcal food poisoning due to low-fat milk in Japan: estimation of enterotoxin A in the incriminated milk and powdered skim milk
.
Epidemiol. Infect
.
130
:
33
40
.
3.
Buchanan
,
R. L.
,
J. L.
Smith
,
C.
McColgan
,
B. S.
Marmer
,
M.
Golden
, and
B.
Gell
.
1993
.
Response surface models for the effects of temperature, pH, sodium chloride, and sodium nitrite on the aerobic and anaerobic growth of Staphylococcus aureus 196E
.
J. Food Saf
.
13
:
159
175
.
4.
Delignette-Muller
,
M. L.
,
L.
Rosso
, and
J. P.
Flandrois
.
1995
.
Accuracy of microbial growth predictions with square root and polynomial models
.
Int. J. Food Microbiol
.
27
:
139
146
.
5.
Ding
,
T.
,
J.
Wang
,
M. S.
Park
,
C. A.
Hwang
, and
D. H.
Oh
.
2013
.
A probability model for enterotoxin production of Bacillus cereus as a function of pH and temperature
.
J. Food Prot
.
76
:
343
347
.
6.
Donnelly
,
C. B.
,
J. E.
Leslie
, and
L. A.
Black
.
1968
.
Production of enterotoxin A in milk
.
Appl. Microbiol
.
16
:
917
924
.
7.
Fujikawa
,
H.
, and
S.
Morozumi
.
2006
.
Modeling Staphylococcus aureus growth and enterotoxin production in milk
.
Food Microbiol
.
23
:
260
267
.
8.
Gibson
,
A. M.
,
J.
Baranyi
,
I.
Pitt
,
M. J.
Eyles
, and
T. A.
Roberts
.
1994
.
Predicting fungal growth: the effect of water activity on Aspergillus flavus and related species
.
Int. J. Food Microbiol
.
23
:
419
431
.
9.
Hennekinne
,
J. A.
,
M. L.
Buyser
, and
S.
Drgacci
.
2012
.
Staphylococcus aureus and its food poisoning toxins: characterization and outbreak investigation
.
FEMS Microbiol. Rev
.
36
:
815
836
.
10.
Huang
,
L. Q.
,
A. J.
Tan
,
Z. W.
Ye
,
L. R.
Zhang
, and
Q. P.
Zhang
.
2009
.
An epidemiological investigation of a food poisoning outbreak associated with dairy produce contamination in kindergartens
.
Mod. Prev. Med
.
36
:
2846
2847
.
11.
Joelle
,
C. H.
,
K. W.
Carl
, and
S. C.
James
.
2009
.
Quantitative microbial risk assessment for Staphylococcus aureus and staphylococcus enterotoxin A in raw milk
.
J. Food Prot
.
72
:
1641
1653
.
12.
Le Loir
,
Y.
,
F.
Baron
, and
M.
Gautier
.
2003
.
Staphylococcus aureus and food poisoning
.
Genet. Mol. Res
.
2
:
63
76
.
13.
Lotter
,
L. P.
, and
L.
Leistner
.
1977
.
Minimal water activity for enterotoxin A production and growth of Staphylococcus aureus
.
Appl. Environ. Microbiol
.
36
:
377
380
.
14.
Mao
,
X. D.
,
J. F.
Hu
, and
X. M.
Liu
.
2010
.
Epidemiologic analysis of 1060 foodborne disease outbreaks from 2003 to 2007 in China
.
China J. Hyg. Res
.
22
:
224
228
.
15.
Notermans
,
S.
, and
C. J.
Heuvelman
.
1983
.
Combined effect of water activity, pH and sub-optimal temperature on growth and enterotoxin production of Staphylococcus aureus
.
J. Food Sci
.
48
:
1832
1840
.
16.
Quigley
,
L. O.
,
O'Sullivan
,
C. Stanton
,
T. P.
Beresford
,
R. P.
Ross
,
G. F.
Fitzgerald
, and
P. D.
Cotter
.
2013
.
The complex microbiota of raw milk
.
FEMS Microbiol. Rev
.
37
:
664
698
.
17.
Rajkovic
,
A.
,
B.
El Moualij
,
M.
Uyttendaele
,
P.
Brolet
,
W.
Zorzi
,
E.
Heinen
,
E.
Foubert
, and
J.
Debever
.
2006
.
Immunoquantitative real-time PCR for detection and quantification of Staphylococcus aureus enterotoxin B in foods
.
Appl. Environ. Microbiol
.
72
:
6593
6599
.
18.
Rho
,
M. J.
, and
D. W.
Schaffner
.
2007
.
Microbial risk assessment of staphylococcal food poisoning in Korean kimbab
.
Int. J. Food Microbiol
.
116
:
332
338
.
19.
Rosengren
,
A.
,
M.
Lindblad
, and
R.
Lindqvist
.
2013
.
The effect of undissociated lactic acid on Staphylococcus aureus growth and enterotoxin A production
.
Int. J. Food Microbiol
.
162
:
159
166
.
20.
Schelin
,
J.
,
N.
Wallin-Carlquist
,
M. T.
Cohn
,
R.
Lindqvist
,
G. C.
Barker
, and
P.
Radstrom
.
2011
.
The formation of Staphylococcus aureus enterotoxin in food environments and advances in risk assessment
.
Virulence
2
:
580
592
.
21.
te-Giffel
,
M. C.
, and
M. H.
Zwietering
.
1999
.
Validation of predictive models describing the growth of Listeria monocytogenes
.
Int. J. Food Microbiol
.
46
:
135
149
.
22.
Tienungoon
,
S.
,
D. A.
Ratkowsky
,
T. A.
McMeekin
, and
T.
Ross
.
2000
.
Growth limits of Listeria monocytogenes as a function of temperature, pH, NaCl, and lactic acid
.
Appl. Environ. Microbiol
.
66
:
4979
4987
.
23.
Tsutsuura
,
S.
,
Y.
Shimamura
, and
M.
Murata
.
2013
.
Temperature dependence of the production of staphylococcal enterotoxin A by Staphylococcus aureus
.
Biosci. Biotech. Biochem
.
77
:
30
37
.
24.
U.S. Food and Drug Administration
.
1992
.
BBB—Staphylococcus aureus. Bad bug book: foodborne pathogenic microorganisms and natural toxins handbook
. .
25.
Valero
,
A.
,
F.
Pérez-Rodríguez
,
E.
Carrasco
,
J. M.
Fuentes-Alventosa
,
R. M.
Garcia-Gimeno
, and
G.
Zurera
.
2009
.
Modelling the growth boundaries of Staphylococcus aureus: effect of temperature, pH and water activity
.
Int. J. Food Microbiol
.
133
:
186
194
.
26.
Vermeulen
,
A.
,
F.
Devlieghere
,
K.
Bernaerts
,
J. V.
Impe
, and
J.
Debevere
.
2007
.
Growth/no growth models describing the influence of pH, lactic and acetic acid on lactic acid bacteria developed to determine the stability of acidified sauces
.
Int. J. Food Microbiol
.
119
:
258
269
.
27.
Yu
,
X. J.
,
J.
Yan
,
J. F.
Zhang
,
C. G.
Xue
,
R.
Dong
,
P. H.
Xie
, and
X. H.
Zhang
.
2010
.
Risk assessment of Staphylococcus aureus in raw milk and establishment of prevention and control measures
.
China Dairy Ind
.
38
:
53
58
.
28.
Zhong
,
T. H.
,
H. Z.
Liu
,
H.
Zhang
, and
L.
Deng
.
2011
.
Investigation of a Staphylococcus aureus food poisoning
.
J. Prev. Med. Inf
.
27
:
735
736
.