Contamination of Clostridium perfringens in soy sauce, and quantitative microbial risk assessment for C. perfringens through soy sauce consumption

Abstract The objective of this study was to conduct QMRA (quantitative microbial risk assessment) of Clostridium perfringens through soy sauce consumption. Four hundred and ninety soy sauce samples from markets were analyzed to detect C. perfringens. Temperature and time were also measured during transportation and display of soy sauce. A primary model was developed by fitting the Weibull model to the C. perfringens cell counts in soy sauce at 7–35°C, and δ (the time needed to decrease 1 log CFU/ml) and ρ (curve shape) were calculated. The parameters were analyzed, using the Exponential model (secondary model) as a function of temperature. The consumption amount and percentage of soy sauce were surveyed, and a dose‐response model was searched. Using all collected data, a simulation model was prepared in the @RISK program to estimate the probability of C. perfringens foodborne illness by soy sauce consumption. C. perfringens were negative in 490 samples. Thus, the initial contamination level was estimated to be −2.9 log CFU/ml. The developed predictive models showed that C. perfringens cell counts decreased during transportation and display. The average consumption amounts, and the percentage of soy sauce were 7.81 ml and 81.2%, respectively. The simulation showed that the probability of C. perfringens foodborne illness by consumption of soy sauce was 1.7 × 10–16 per person per day. Therefore, the risk of C. perfringens by consumption of soy sauce is low in Korea.

raised. C. perfringens is one of the most common causes of foodborne disease outbreak in the United States and Korea, alongside norovirus, Salmonella, Campylobacter, and Escherichia coli (Centers for Disease Control & Prevention, 2017;Ministry of Food & Drug Safety, 2020).
Clostridium perfringens are anaerobic bacteria and exist as vegetative cells or spores in food. Cooking (60°C) kill the vegetative cells, but the spores. These spores can germinate into vegetative cells at 20-30°C, which is the temperature for soy sauce storage, and the vegetative cells produce toxin in the intestinal tract (McClane et al., 2012). Hence, the contamination of C. perfringens in soy sauce and the risk of C. perfringens by soy sauce consumption should be evaluated.
Quantitative microbial risk assessment (QMRA) is a scientific evaluation method for estimating risks caused by microorganisms, including hazard identification, exposure assessment, hazard characterization, and risk characterization. Thus, to estimate the risk of C. perfringens contamination in food, QMRA is appropriate (Codex Alimentarius Commission, 2015;Jeong et al., 2018;Lee et al., 2019;WHO, 2002). Also, risk assessment of C. perfringens should be assessed through risk assessment based on predictive models. Therefore, the objectives of this study were to investigate the contamination of C. perfringens in soy sauce and to evaluate the risk of the pathogen through soy sauce consumption.

| Investigation of C. perfringens contamination in soy sauce
Four hundred and ninety soy sauce samples were purchased from supermarkets in Korea. One milliliter of the soy sauce was spreadplated on tryptose sulfite cycloserine agar (TSC; Difco) plates and incubated at 37°C for 18-24 hr under the anaerobic condition created by an anaerobic pack (Oxoid; Basingstoke) in anaerobic jar. The resulting colonies were analyzed, using a compact VITEK ® 2 ANC ID card (BioMérieux). When the result indicated acceptable or higher, it was determined as positive for C. perfringens. The prevalence data of C. perfringens were fitted to the Beta distribution (α 1 , the number of positive samples + 1; α 2, tested total samples-positive samples + 1), and the initial contamination level (CFU/ml) was then evaluated, using the equation: [-LN(1-Beta(α 1 , α 2 ))/ml] (Sanaa et al., 2004;Vose, 1998).

| Development of predictive models
Clostridium perfringens strains KCCM (Korean Culture Center of Microorganisms) 12098, KCCM40946, KCCM40947, and KCTC (Korean Collection for Type Cultures) 5101 were cultured in 10 ml cooked meat medium (CMM; Oxoid, Basingstoke, Hampshire, UK) at 37°C for 24 hr under the anaerobic conditions. Three-milliliter aliquots of the cultures were inoculated in 30 ml Duncan and strong medium (15.0 g proteose peptone, 10.0 g sodium phosphate, 4.0 g raffinose, 4.0 g yeast extract, 1.0 g sodium thioglycolate, and 50.0 ml 0.51 mM caffeine in 1 L distilled water) and anaerobically incubated at 37°C for 48 hr. The C. perfringens strains were harvested by centrifugation at 1,912 × g and 4°C for 15 min. The supernatant was discarded, and the cells were washed twice with phosphate buffered saline (PBS; pH 7.4, 8.0 g NaCl, 1.5 g NaHPO 4 , 0.2 g KH 2 PO 4 , and 0.2 g KCl in 1 L distilled water).
The cell pellets were resuspended in 3 ml PBS, and the strains were mixed to be used as the inoculum. The C. perfringens inoculum was inoculated to 10 ml soy sauce in a conical tube to achieve a final concentration of 3.0 log CFU/ml, and stored at 7, 15, 25, and 35°C for up to 5 days. During storage, 10 ml of the soy sauce was diluted with 20 ml buffered peptone water (BPW; Dickinson and Company), and it was vortexed. The homogenates were spreadplated on TSC agar, and the plates were incubated anaerobically at 37°C for 24 hr. Typical black colonies with halos were counted, and the C. perfringens cell count data were fitted to the Weibull model.

The equation was as follows;
where N is cell counts, N 0 is initial cell counts, δ is treatment time for the first decimal reduction and ρ is curve shape (Mafart et al., 2002).
To describe the effect of temperature on δ, a secondary model was developed with the equation as follows; Where α is constant, k is the rate constant and T is storage temperature. To evaluate the model performance, C. perfringens was inoculated in soy sauce as described above, and the samples were stored at 20 and 30°C. During storage, C. perfringens cells were counted as described above to obtain the observed values. The differences between the observed data and predicted data derived from the developed predictive models for 20 and 30°C were quantified by calculating the root mean square error (RMSE) values, bias factor (B f ), and accuracy factor (A f ) as follows ; where n represents the number of data points. The RMSE is the square root of the variance of the residuals. It indicates the absolute fit of the model to the data-how close the observed data points are to the predicted values. The B f shows by how much models over-predict (B f > 1) or underpredict (B f < 1), and A f shows by how much the predicted data differ from the observed data (Yoon et al., 2011).

| Investigations for soy sauce storage conditions and consumption amounts
The temperature of the transport vehicle was collected by measuring the internal temperature of the transport vehicle of four companies in a retail market. Storage temperature and time data for soy sauce at retails were collected through interviews with managers at retail markets.
The transport temperature data were analyzed by @RISK 7.6 (Palisade Corp.) to determine an appropriate probabilistic distribution. To have the daily consumption amounts and percentage for the soy sauce.
The data from the Korea National Health and Nutrition Examination Survey 2017 were used. The consumption data were analyzed, using the @RISK 7.6 to determine an appropriate probabilistic distribution.
The consumption percentage of soy sauce was calculated by dividing the number of total respondents (6,628 people) in the survey to the number of respondents (5,382 people) who consumed soy sauce.
F I G U R E 1 Cumulative probability of simulated initial contamination level of Clostridium perfringens in soy sauce F I G U R E 2 Fitted for Clostridium perfringens cell counts in soy sauce with primary models during storage at 7, 15, 25, and 35°C

| Search for dose response model
To evaluate the response of a human to C. perfringens dose, when people intake soy sauce contaminated with C. perfringens consumption, dose response models were searched through published articles.

| Prevalence and initial contamination levels of C. perfringens
Of 490 samples, C. perfringens cell counts were below the detection limit (1 log CFU/ml) in all samples. Because there were no C.
perfringens positive samples, contamination levels need to be estimated. Thus, the Beta distribution was used to estimate the initial contamination level of C. perfringens in soy sauce (Vose, 1998), and the Beta distribution (RiskBeta [1,491]) was used to, and initial C.
perfringens contamination levels were estimated to be −2.9 log CFU/ ml (Figure 1).

| Predictive model to describe the kinetic behavior of C. perfringens
To describe the C. perfringens cell count changes in soy sauce dur- predictive models were appropriate to predict the C. perfringens cell counts in soy sauce during transportation and storage.

| Amounts and percentage of soy sauce consumption
Raw data for daily consumption amounts of soy sauce extracted from

| Risk characterization
Employing estimated C. perfringens contamination level, predictive models simulating C. perfringens cell counts with the probabilistic distributions of temperature and time, probabilistic distribution of consumption amounts, consumption percentage, and dose response model, the simulation model was prepared as shown in Table 2. The cumulative density determined by this simulation showed that estimated C. perfringens cell counts decreased gradually from initial contamination (IC) to display (C3) (Figure 4). The simulation result showed that the risk of C. perfringens per person per day through soy sauce consumption was 1.7 × 10 −16 in Korea (Table 3). Also, the risk has a positive correlation with prevalence, consumption amount, and consumption percentage. On the other hand, it shows a negative correlation with market display time and market storage time. (Figure 5).
In a study by Ko et al. (2012), the probabilities of C. perfringens foodborne illness per person per day for ham and sausage products was estimated as 3.97 × 10 −11 ± 1.80 × 10 −9 . Also, the probability of C.
perfringens foodborne illness by cheese consumption for natural and processed cheeses was 9.57 × 10 −14 and 3.58 × 10 −14 per person per day, respectively (Lee et al., 2016). Comparing these risks with the risk estimated in our study the risk of C. perfringens foodborne illness from soy sauce consumption was relatively low. This result might be caused by high salt concentration of soy sauce, which decrease C. perfringens cell counts are suggested in many studies (Chhetri & Karki, 2014;Gibson & Roberts, 1986;Juneja & Marmer, 1996).

| CON CLUS IONS
In conclusion, the risk for C. perfringens foodborne illness soy sauce consumption in Korea is low, because although soy sauce is consumed widely, C. perfringens prevalence in soy sauce is low, and C. perfringens cell counts decreased during transportation, storage, and display.

ACK N OWLED G M ENTS
This research was supported by a grant (17162KFDA035) from the Ministry of Food and Drug Safety in 2017.

F I G U R E 5
Correlation coefficients for risk factors affecting the probability of Clostridium perfringens illness per person per day by soy sauce consumption