Foodborne diseases have been, and remain, a major global challenge in public health and economic development, with increasing numbers of incidents recorded in many countries worldwide. The WHO study on the global burden of foodborne disease estimates that 31 foodborne hazards cause 600 million illnesses and 420,000 deaths (WHO, 2015). Efforts are being made by all parties throughout the entire food chain to prevent, detect, and manage hazards resulting from the presence of foodborne pathogens. Steady changes in consumers’ eating patterns and preferences along with increasing health concerns and environmental awareness have led to a continuous and growing demand for, e.g., ready-to-eat meals, minimally processed foods, and local products (Lupien, 2007; Lappo et al., 2015). New challenges are thus relentlessly arising in the battle against foodborne diseases. This is particularly the case for minimally processed foods that provide favorable environments for most pathogenic bacteria, among them Staphylococcus aureus. Staphylococcal food poisoning (SFP) is a foodborne intoxication caused by staphylococcal enterotoxins (SEs) and not the bacteria themselves. Thus, the virulence of S. aureus as a foodborne pathogen depends on the amount of SEs formed.

Proof of the association of S. aureus with food poisoning was obtained in 1914 by Barber (1914) who demonstrated that food poisoning resulted from the consumption of unrefrigerated milk from a cow with staphylococcal mastitis. Dack et al. (1930) showed that a filterable toxin, later called enterotoxin, was the causative agent of SFP. Today, S. aureus is considered one of the most common pathogens causing food intoxication (Fetsch and Johler, 2018). In the United States, approximately 241,000 cases of foodborne illness were reported to have been caused by S. aureus between 2000 and 2008, placing the pathogen in the fifth place among those most commonly reported (Scallan et al., 2011). In the European Union (EU), SEs were reported to be the causes of 8% of foodborne outbreaks in 2016, rendering bacterial toxins the third most common causative agent of outbreaks (EFSA, 2017). In other parts of the world, such as Australia, intoxication by S. aureus accounted for 1% of all confirmed and suspected foodborne outbreaks between January 2000 and March 2012 (Pillsbury et al., 2013) whereas in China it was reported in 2013 that 12.5% of all foodborne bacterial outbreaks were caused by S. aureus (Wang et al., 2017). S. aureus is found on the skin and mucous membranes of humans and other warm-blooded animals, and is widespread in the environment. It can enter the food chain through various routes, the most common being via raw materials, food handlers, or poor hygiene in food processing equipment (Hennekinne et al., 2012; Fisher et al., 2018). In the case of S. aureus, contamination during food preparation and processing, or during post-production, is regarded as the main sources of risk (Argudín et al., 2010; Hennekinne et al., 2010; Kadariya et al., 2014; Castro et al., 2016). Ready-to-eat foods containing meat products and dressings, which are handled and consumed without further treatment, have been implicated in a number of SFP outbreaks (Table 1). In the EU summary report 2016 (EFSA, 2017) on foodborne outbreaks, the main vehicle for SFP outbreaks was reported to be “mixed foods” with 31% followed by milk and milk products with 22%, highlighting the impact of post-processing of foods on the development of the disease.

Staphylococcus aureus is able to grow and produce enterotoxins under a wide range of temperatures, pH values, water activity (a_w) levels, and sodium chloride concentrations (Le Loir et al., 2003; Paulin et al., 2012; Adams et al., 2016). This robustness of the pathogen in a wide range of environmental conditions broadens the diversity of foods in which S. aureus can grow and express virulence (Hennekinne et al., 2012; EFSA, 2017). Thus, any food product that can support the growth of S. aureus could pose a risk in terms of SFP. Additional post-production handling of the food could further increase the risk of disease due to S. aureus contamination. Identifying the risks associated with S. aureus in the food handling and production environment, such as breach of hygiene, is one of the most important factors in the control of SFP. However, the dissemination of the pathogen in nature, its robustness under diverse conditions, and the level of knowledge of people involved in food processing are only some of the challenges associated with this pathogen. To these must be added specific differences among S. aureus strains including the variation in SE expression and production levels, the different types of mobile genetic elements encoding the se genes, and the accompanying range of different regulatory mechanisms. As stated above, SFP leads to intoxication due to the ingestion of enterotoxins preformed in the food (Fetsch and Johler, 2018). SEs produced by S. aureus are resistant to most of the treatments that eliminate bacterial cells, such as heat treatment and low pH, and can therefore be present in food even after processing. Their resistance to proteolytic enzymes also enables them to pass through the digestive tract unaffected, thus increasing the risk of illness (Le Loir et al., 2003). The key element in preventing SFP outbreaks is thus to control the production of enterotoxins.

Among the SEs and SE-like proteins produced by S. aureus, those usually associated with SFP are the five classical SEs: SEA–SEE (Wieneke et al., 1993; Cha et al., 2006; Kérouanton et al., 2007; Rajkovic, 2012). Staphylococcal enterotoxin A (SEA) is most frequently reported as the causative agent of SFP outbreaks (80% of cases), followed by SEB, SEC, and SED (Pinchuk et al., 2010; Hennekinne et al., 2012; Sospedra et al., 2013). The challenge in understanding and controlling the production of SEs lies in the complexity and diversity of the encoding genetic elements and regulation mechanisms. The se genes are carried on various genetic elements such as plasmids, bacteriophages, pathogenicity islands (SaPIs), and the enterotoxin gene cluster (egc) (Genigeorgis, 1989; Le Loir et al., 2003; Argudín et al., 2010; Hennekinne et al., 2012). Different regulatory mechanisms are involved in SE expression depending on the genetic element carrying the SE.

The mechanisms controlling SE production in S. aureus are numerous and include se gene promoter regions, multiple global regulators of virulence, such as the accessory gene regulator (agr), the staphylococcal accessory regulator (sar), the repressor of toxins (rot), the sigB factor, and the two-component system S. aureus exoprotein expression (sae), in addition to a cytoplasmic SEB form that controls SEB production (Gustafson and Wilkinson, 2005). Selected classical enterotoxins with respective regulatory mechanism/s/ are summarized in Table 2 and also well accompanied in Figure 1 in the review by Fisher et al. (2018) that gives an overview of the relationship between growth environment/condition, regulatory mechanisms, and type of SE (Fisher et al., 2018). Knowledge concerning the regulatory mechanism of a SE can therefore be of great value in the effective prevention of S. aureus intoxication, as the appropriate interventions may differ depending on the SE-producing S. aureus strain. SEA and SED can be used to exemplify different mobile elements and different regulatory systems, as SEA is encoded by a bacteriophage, while SED is encoded by a plasmid. The latter is partly regulated by the agr system, a quorum sensing system that enables S. aureus to respond to cell density, and SED is thus mainly produced during the transition from the exponential to the stationary growth phase of the microorganism (Schelin et al., 2011). The regulation of SEA, however, has been linked to the life cycle of the bacteriophages carrying the sea gene, with SEA production being the highest during the exponential growth phase, at the peak of replication (Derzelle et al., 2009; Schelin et al., 2011). Furthermore, these regulatory mechanisms are affected in different ways by the environmental conditions in a food product, such as salt content, water activity, and pH (Schelin et al., 2017). This can have a further impact on the production of the respective enterotoxins, making them difficult to control.

The location of the se genes on mobile genetic elements presents an additional challenge in SFP control, as it supports horizontal gene transfer between different S. aureus strains, and therefore the dissemination of se genes (Hennekinne et al., 2012). The transfer of genetic elements in S. aureus has contributed to strain variability and enhanced virulence, as S. aureus strains usually carry more than one se gene (Becker et al., 2003). These evolutionary trends in S. aureus call for constant research to identify potentially new strains and to understand their behavior and virulence expression with regard to SFP.

The diversity of food products implicated in SFP is apparent from documented outbreaks (Table 1). The robustness of S. aureus under a wide range of environmental conditions should be considered a virulence parameter, making it difficult to predict the general behavior of the pathogen in various food matrices under different environmental conditions. The series of events in an SFP outbreak typically include the following: (i) the presence of the pathogen in raw materials or the food handler(s), (ii) contamination of the food, for example, through processing equipment or through the food preparer, (iii) inappropriate storage conditions and/or inadequate temperature control that allow for bacterial growth and enterotoxin production, and (iv) ingestion of contaminated food containing a sufficient amount of SE to trigger symptoms of the disease. The majority of well-documented outbreaks indicate that the most common means of contamination in cases of SFP are poor hygiene practices during the processing, cooking, and distribution of food products. Inadequate cooling of food products is the main cause of S. aureus growth leading to disease (Pereira et al., 1996; Asao et al., 2003; Schmid et al., 2009; Hennekinne et al., 2012).

One of the first well-documented SFP outbreaks was reported by Denison (1936) and involved the consumption of contaminated cream puffs by high school students. In this case, the source of contamination was not actually identified. In more recent years, a large-scale SFP outbreak affected as many as 13,420 individuals, and involved the consumption of dairy products that contained SEA (Asao et al., 2003). From the investigation of this outbreak it was found that the source of contamination was the powdered skim milk used to produce the dairy products involved in the outbreak. The most interesting factor in this case was that the contaminated products had been heat-treated, in some cases as much as three times, at 130°C for 2 or 4 s, and bacterial cells of S. aureus were completely eliminated. SEA retained, however, both its immunological and emetic activity and led to food poisoning (Asao et al., 2003). Besides the classical types of food products associated with S. aureus contamination an example of a more recent study investigated the prevalence on various retail vegetables in China between 2011 and 2016 (Wu et al., 2018). Although the levels of S. aureus were moderate, it was found in 5.73% of the samples and lettuce was the most common vegetable. In addition to the conventional entry routes connected with production and processing environments, the global trade and especially the illegal transportation of food across borders have also been found to contribute to the transmission of S. aureus (Müller et al., 2016; Rodriguez-Lazaro et al., 2017). Another alarming concern is that not only enterotoxigenic but also methicillin-resistant S. aureus (MRSA) strains are increasingly being isolated from a variety of meat and dairy-based products as well as ready-to-eat foods indicating that the food chain is one additional and alternative reservoir as well as transmission route for antimicrobial resistant bacteria (Sergelidis and Angelidis, 2017; Islam et al., 2019; Wu et al., 2019).

Staphylococcal enterotoxin A is one of the more extensively studied SEs, mainly due to the fact that it accounts for 80% of reported SFP cases. Its worldwide predominance has been extensively documented (Argudín et al., 2010). In a study of 359 SFP outbreaks in the United Kingdom that took place between 1969 and 1990, it was shown that 79% of the S. aureus strains involved produced SEA (Wieneke et al., 1993). SEA was the only enterotoxin detected in 56.9% of the outbreaks, while in a few cases it was detected together with SED, SEB, and SEC, or with SEB and SED. The sources implicated included meat, poultry, and products thereof, especially ham and chicken. The sea gene was also the predominant enterotoxin gene detected in dairy products implicated in SFP outbreaks in Brazil (Veras et al., 2008). SEA was the most frequent causative agent in 69.7% of SFP outbreaks in France between 1981 and 2002 (Kérouanton et al., 2007). A challenging aspect of this high prevalence with regard to controlling SFP intoxication is that SEA is one of the few SEs whose regulation is not dependent on the agr system. A number of studies have investigated the expression of the sea gene in connection with alternative regulatory mechanisms, in order to understand the conditions under which the enterotoxin is produced. In the following section, SEA will be used as a characteristic example of the challenges from different genetic backbones; the regulatory mechanisms; and environmental stress responses to the robustness, growth behavior, and strain variations associated with SE production by S. aureus leading to SFP.

Describing and quantifying bacterial growth in food environments are rather complex – as complex is the response of the bacteria in the food matrix itself. Furthermore, changes in food production methods and pathogen evolution create a need for continuous research on microbial risk assessment. Understanding the genetic mechanisms that regulate the phenotypic responses of bacteria, and incorporating this knowledge into existing predictive models, will greatly improve hazard identification and pathogen control. Risk assessment associated with SFP has another degree of complexity due to the necessity of assessing SE production, and not just the presence or absence of the organism. This means that the strain of S. aureus, the type and amount of SE produced, and the possible correlation between growth and SE production must be evaluated. The potential for intoxicating amounts of SE to be present in a food product, even in the absence of viable S. aureus cells, should also be considered when assessing the risk posed by this organism (Cretenet et al., 2011).

A number of studies have been carried out, with focus on assessment of S. aureus risk of growth and toxin production in various products, including milk, unripened raw-milk cheese, home-cooked foods, and cream-filled bakery products (Lindqvist et al., 2002; Stewart et al., 2003; Taulo et al., 2008; Heidinger et al., 2009; Kim et al., 2009). As one of the major components of risk assessment is the exposure assessment, the use of predictive models is an essential tool for estimating the spatiotemporal changes of S. aureus population in the food chain, the attainment of toxin producing levels in target foods, and the amounts of toxins present in foods at the time of consumption. The existing predictive models are capable of simulating dynamics of cell populations and toxin production over time, in response to the main factors, controlling microbial growth, such as temperature, pH, and water activity. Both kinetic and probabilistic models are available, as detailed in the following lines, some also being readily available in predictive modeling databases, such as ComBase. Nonetheless, the polynomial models of ComBase are generic models trained on responses of multiple S. aureus strains to temperature (7.5–30°C), pH (4.4–7.0), and a_w (0.907–1) in broth. As a counteract to the limitations of broth-based models, growth simulations via the above predictive modeling platforms enable the users to assess the impact of variability in population dynamics, thereby assessing the stochastic response of microorganisms, which is an important feature for exposure assessment.

Regarding the food-specific models, Fujikawa and Morozumi (2006) developed a model describing the rate of SEA production on the basis that the enterotoxin is formed when S. aureus levels are greater than 6.5 log cfu ml^-1, and that the relationship between SEA production and growth is linear in the temperature range 15–32°C. In a study by Valero et al. (2009), the influence of temperature, pH, and water activity on the growth of S. aureus was evaluated and fitted in a growth/no growth model. Lindqvist et al. (2002) used predictive modeling and survey data in combination with probabilistic modeling to simulate the levels of S. aureus at the time of consumption of unripened cheese made from raw milk. In this way, they evaluated the risk associated with consumption of this cheese and found that the initial S. aureus population, pH, and storage time were the main risk factors. Barker and Goméz-Tomé (2013) took risk assessment one step further and developed a probabilistic model for the representation of the risks that arise in pasteurized milk from the presence of S. aureus, and particularly SEs, during the entire production chain. Specifically, they implemented probabilistic analysis with a Bayesian belief network, and in that way introduced the concept of biotraceability. Thus, the model developed by Barker and Goméz-Tomé (2013) allowed the identification of the main hazard sources. This in turn allowed for conclusions on where pre-emptive action should be taken in the milk production process to increase its efficiency and safety (Barker and Goméz-Tomé, 2013).

A limitation of the models developed to date, such as those described above, is the fact that the risk is assessed based on predicted levels of S. aureus that have been associated with enterotoxin production, rather than predicted enterotoxin levels in the final product, or the actual dose–response. Moreover, the information usually included in these models is generic, i.e., it does not distinguish between the different types of enterotoxins and their gene regulatory mechanisms. The latter is a key challenge for assessing the severity of the disease and for characterizing the impact of the food (micro)-environment and storage conditions on the ability of S. aureus to produce toxin. In this context, the marginal conditions allowing production of SEA are different from those that permit growth of S. aureus (Borneman et al., 2009; Ding et al., 2016), whereas toxin production may also be induced by severe stresses (Cao et al., 2012). Nonetheless, despite the existing preliminary indications, whether and how frequently these types of stress are associated with food environments requires further investigation (Zeaki et al., 2014, 2015b). Overall, the aforementioned evidence suggests that the toxin producing S. aureus levels are practically variable and a generic threshold cannot be reliably set. Usually in quantitative microbial risk assessment (QMRA) approaches, the prediction of S. aureus growth by the time of consumption is linked to the correlation between levels of S. aureus and toxin and the assumption of a minimum toxin amount of 20–50 ng for the occurrence of intoxication symptoms, in order to assess the risk of intoxication (Kim et al., 2009).

Many of the existing studies highlight the need for more information regarding the dynamics of S. aureus growth in relation to gene expression and enterotoxin production, the impact of different food environments on gene regulation, and relevant dose–response data, to improve risk assessment (Lindqvist et al., 2002; Fujikawa and Morozumi, 2006; Schelin et al., 2011; Barker and Goméz-Tomé, 2013). Rosengren et al. (2010) addressed the knowledge gaps in risk assessment associated with S. aureus in fresh and short-time ripened cheeses in Sweden. The areas where gaps were identified included producers’ practices at farm dairies, the properties and the virulence of various S. aureus strains, the sources of S. aureus, and the impact of pasteurization and starter cultures on S. aureus levels. Given that S. aureus is considered a rather “poor” competitor, predictive models have also quantified the competitive effect of starter cultures (e.g., lactic acid bacteria) on the growth of S. aureus, in response to temperature (Le Marc et al., 2009). The model was based on the so-called “Jameson effect,” which practically describes the growth cessation of the weaker organism in a binary culture, when the competitor reaches a threshold population, which is higher than the other organisms.

The complexity of SE production and the network of synergistic signals that regulate enterotoxin expression highlight the need for physiology-based, or even genotype-specific S. aureus risk assessments, thus, paving the way of increasingly integrating -omics in QMRA. These issues are further discussed in the following lines. The S. aureus strain and, in the case of phage-encoded enterotoxin genes such as sea, the respective sea-carrying phage are critical parameters influencing the relevance of S. aureus as a potential etiological agent in an SFP outbreak and need to be taken in account for risk assessment. Moreover, the effect of the environment on enterotoxin expression should be considered when assessing the risk of SFP. Various studies evidence the limitation on the reliability of risk assessment based on the assumption that cell growth and enterotoxin production are related (Marta et al., 2011; Zeaki et al., 2014; Schelin et al., 2017). In addition to what was mentioned above for the probability of toxin production it is also the S. aureus growth rate that follows different patterns from those of enterotoxin production, depending on growth conditions and strain. For SEA, more studies, including different S. aureus food isolates carrying different Siphoviridae bacteriophages could further improve our knowledge concerning the regulation of expression. The schematic representation of the events regulating prophage induction in the study by Zeaki et al. (2015a) could serve as a template into which more information on sea gene expression could be incorporated to create a model including the factors that should be considered during risk assessment for SFP (Figure 2). Furthermore, additional data on the growth of S. aureus and SEA production on various food matrices could improve quantitative models for more accurate estimates of SEA production.

Another challenge that remains to be overcome when assessing the risk of S. aureus intoxication is the number of se genes carried by one strain. As shown by relevant studies and SFP outbreaks, it is quite common for more than one se gene to be detected in S. aureus strains implicated in intoxication cases (see e.g., Hennekinne et al., 2009 [sea, sed, sej]; Kitamoto et al., 2009 [sea, sed]; Schmid et al., 2009 [sea, sed]; and Johler et al., 2015 [sea, sed] in Table 1). One of the most recent examples is an outbreak documented by Johler et al. (2015), which occurred in a Swiss boarding school and affected 14 individuals. The S. aureus strain identified as the source of that outbreak harbored both sea and sed. Likewise, both SEA and SED were recovered from the implicated food (soft cheese) at levels of >6 ng/g cheese and >200 ng/g cheese, respectively. Since SEA and SED production are regulated by different mechanisms, it is evident that more information is needed on how these mechanisms are triggered in different food environments.

Identifying the point of entry of the S. aureus strain into the food chain and the clonal lineage to which it belongs will provide information that could be used in risk assessment profiles for identification of relevant preventive controls. A recent study by Kümmel et al. (2016) provides insight in this matter. Through the collection of samples from cow to retail product (ripened semi-hard cheese) at 18 dairy farms, the authors found that the bovine udder constitutes an important source of S. aureus in the dairy industry. Strains of genotype B were those most successfully transmitted from cows into the dairy production chain. The strain implicated in the outbreak at the Swiss boarding school was also assigned to an S. aureus of genotype B (Johler et al., 2015). Similarly, the study by Hummerjohann et al. (2014) showed that genotype B strains are primarily found in semi-hard cheeses produced from raw milk.

It should be borne in mind that pathogenic bacteria evolve with consumers’ changing habits and the development of food production methods, and that microbial risk assessment should also develop in a similar manner. A typical example of studies in this direction is the one assessing the impact of temperature fluctuations in toxin production of S. aureus in deli meals (Røssvoll et al., 2014). Moreover, the recognition that bacterial responses are the result of a complex network of genetic events should direct research efforts toward a better understanding of these mechanisms, with the aim of more efficient pathogen control throughout the food production chain.

Staphylococcus aureus is a highly evolved pathogen that brings about a number of challenges to food production. The physiological properties of the bacterium as such pose the first challenge for SFP containment. S. aureus robustness for growth along with its natural niches assists in the establishment of the pathogen in a broad range of food products. Poor hygiene habits by food handlers further facilitate its entrance into the food chain. In addition, S. aureus is an evolutionary flexible pathogen. Horizontal gene transfer is one of the main ways this bacterium acquires virulence factors. Therefore, the continual changes in food habits and processing methods trigger the adaptation mechanisms of the pathogen. The latter has brought about new strains exhibiting increased virulence and resistance to the applicable preservation methods. The variety of SEs produced by S. aureus and the complexity of their regulation and production must moreover be carefully considered. Food parameters can critically affect SE production and hence the virulence of S. aureus. Understanding the mechanisms behind the expression of an SE and how they can be influenced by for example the food composition, salt content, and pH constitutes the greatest of the challenge to overcome in the battle against SFP. To conclude, more targeted studies of different SE regulatory mechanisms in food will be the key to allow for improved understanding of SFP and to enable risk assessment. Finally, considering the exponential advances in the area of -omics disciplines, an ultimate challenge and emerging trend in food science is to integrate the molecular data underpinning the phenotypic responses of bacterial pathogens into predictive models and then into QMRA. This will enable our knowledge on the physiological response of pathogens, which may explain uncertain areas of their behavior (at least currently) to be expressed in quantitative terms and translated in explicit (numerical) terms of risk for used in proper decisions to guard food safety.

NZ designed this review, performed the literature study, prepared Table 1 and Figure 1, 2, and wrote the major part of the manuscript. SJ and JS prepared Table 2. SJ, PS, and JS contributed in writing and critical revision. All authors approved the final version of the manuscript. JS was the principal supervisor and responsible for manuscript preparation.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.