Fresh pasta filata cheeses have been usually considered among the microbiologically safe cheeses because of the supplementary effect of the stretching of the acid cheese curd with hot or boiling water. Nonetheless, Mozzarella di Bufala Campana (MBC) Protected Designation of Origin (PDO) cheese is characterized by pH (>5.1) and water activity (a_w > 0.975) values that do not allow to automatically consider it as food not able to support Listeria monocytogenes growth, according to Commission Regulation (EC) No. 2073/2005 (2020). From 1998 to 2017 out of more than 370 cases of cheese microbial contamination, only three cases of Mozzarella withdrawn from the market because of the presence of L. monocytogenes have been reported by European Rapid Alert System for Food and Feed (RASFF Portal, 2020). Suddenly, in 2018 and 2019 three recalls of MBC occurred in France (RASFF 2018/1198 and 2018/2459) and in Canada (Canadian Food Inspection Agency (CFIA), 2019), because of the presence of L. monocytogenes. This unusual contamination has urgently raised the need to investigate on its possible reasons, since the risk profile of MBC may be considered lower than the more widely consumed cow milk Mozzarella, because of the higher acidity of the cheese curd (pH 4.8 vs pH 5.1) and the higher temperature of the stretched curd at the end of the process (about 68–72°C vs 58–63°C; Mucchetti and Neviani, 2006). The stretching process of buffalo cheese curd is associated to significant but extremely variable decimal reductions of L. monocytogenes count, ranging from 1 to 8, as resulted by some studies performed with laboratory scale cheese making trials (Villani et al., 1996; Raimundo et al., 2013; Serraino et al., 2013; Murru et al., 2018). The survival probability depends on the cheese curd contamination, the effective conditions of stretching applied during the experiments, often not fully described, and the heat resistance of biotypes. Up to present, at the best of our knowledge, no L. monocytogenes contamination has been reported in MBC cheese curds. Its presence at the moment of stretching is believed to be occasional and lower than the counts measured during challenge tests, where artificial curd contamination ranged from 10³ to 10⁷ CFU/g. In these challenge tests, L. monocytogens was added to milk, and during cheese curd maturing a decrease (Murru et al., 2018) or even a small but significant growth (Serraino et al., 2013) have been observed. The degree of L. monocytogenes survival after stretching may be associated to the stretching conditions, as the overall amount of thermal and mechanical energy applied, that determine the temperature and time of permanence of the cheese curd at lethal temperature, is inversely related to the degree of cheese curd maturity and demineralization. Indeed, to obtain a cheese with the standard and typical MBC structure, the cheesemaker lowers the stretching temperature when the cheese curd has a too low pH and, as a consequence, an excessive extent of casein demineralization (Mucchetti and Neviani, 2006; Mucchetti et al., 2016).

Moreover, further aspects should be considered to explain the variable effects of curd stretching on L. monocytogenes survival. First, heterogeneity (genetic or phenotypic diversity) within a population, can promote adaptation and survival when the population experiences sudden environmental changes (Ryall et al., 2012; Davis and Isberg, 2016). These phenomena may explain the ability of a fraction of a bacterial population to withstand stresses that kill the majority of the population itself (Booth, 2002), and could be associated both to Listeria spp. biodiversity (Bernini et al., 2013) and to the presence of isogenic variants, able to influence its heat tolerance, e.g., because of a different ability to repair cell membrane after sub lethal heat injury (Somolinos et al., 2010). Heterogeneity can further contribute to explain the limits of the widely used primary model of Bigelow (Bigelow, 1921), when deviations from linear relationship between time of heat treatment and survival are observed (Cebrian et al., 2017). The heat resistance range of L. monocytogenes biotypes is quite variable, as reported by Doyle et al. (2001) and Aryani et al. (2015b), with D-values at 60°C ranging from few seconds to more than 4 min, and decreases at low pH, high a_w and low NaCl content (Jørgensen et al., 1995; Mazzotta, 2001; Bucur et al., 2018). Heat resistance of L. monocytogenes is also influenced by the growth phase, with the cells in exponential phase being more heat sensitive than those in stationary phase (Lou and Yousef, 1996; Jørgensen et al., 1999; Aryani et al., 2015b), and by stress adaptation phenomena, e.g., acid and or osmotic stress (Jørgensen et al., 1995; De Jesús and Whiting, 2003). Moreover, it may be interesting to note that some D- and z-values measured in some complex food matrices (meat, eggs) were significantly higher than those determined in culture media as Tryptone Soy Broth (TSB; Quintavalla and Campanini, 1991; Doyle et al., 2001), confirming the role of the interaction of microorganisms with the matrix in determining its stress tolerance.

Furthermore, the presence of L. monocytogenes in MBC can also be the result of post-contamination during the processing steps after stretching and before packaging. After stretching, MBC cheese is hardened and cooled by dipping into flowing tap water, salted by immersion in a brine and finally packed by dipping into a covering liquid, composed by salt and organic acids (e.g., lactic acid; Mucchetti and Neviani, 2006). Potentially, the contamination of these fluids can be transferred to the cheese, but at present time no data demonstrated the presence of L. monocytogenes in tap water and/or brines and/or covering liquids used for MBC cheese making. However, even if the presence of L. monocytogenes in tap water is not largely documented (Lyautey et al., 2007), its ability to survive (Budzińska et al., 2012), and to create biofilm (Gião and Keevil, 2014) has been determined. Otherwise, the presence of L. monocytogenes in brines of different cheeses has been detected using analytical tools as quantitative polymerase chain reaction (PCR) (Alessandria et al., 2010; Barancelli et al., 2014), and its ability to survive was inversely related to the acidity and NaCl content of the brine itself (Schirmer et al., 2014). The presence of this pathogen on cheese surface and/or on equipment may result in water and brine contamination, causing the beginning of a non-controlled cross-contamination. Also, the more frequent presence of Listeria spp. on non-food contact surfaces (floor, drains, walls, and platforms) may become a potential route of cheese contamination (Barancelli et al., 2014). To avoid cheese cross-contamination, the treatment of water, brines and covering liquid may become necessary. Beside heat processes, filtration, and/or UV-C treatment (Gayan et al., 2015) can be effective in reducing the count of L. monocytogenes. However, heat treatments can lead to more reproducible results.

Knowledge of decimal reduction time (D-value) and thermal resistance constant (z-value) in each out of these matrices can allow to build primary and secondary models able to foresee the lethal effect of heat treatments, contributing to improve the efficacy and the sustainability of the processes, avoiding to waste energy and/or time, mainly when batch treatments are applied as occur in many small sized MBC dairies. The D-value is the time required to obtain a ten-fold reduction of a microbial population at a constant lethal temperature and it varies according to the microorganism and to the medium where heat is applied. The z-value is the temperature variation required for the D-value to change by a factor of ten. Using these parameters, it is possible to foresee the effect of a thermal treatment, accepting that the death rate follows a first order reaction kinetics. Despite the death rate of many microorganisms does not always or completely follow this order of kinetics, suggesting that other models (e.g., Weibull frequency distribution models or non-log-linear Geeraerd’s model) could be applied (Peleg and Cole, 1998; Mafart et al., 2002; Geeraerd et al., 2005; Cebrian et al., 2017), the linear model (Bigelow, 1921) continues to offer a relevant contribution to predict the effect of a heat treatment. However, in the presence of a non-isothermal temperature profile, as in the case of curd stretching operation, it is necessary to develop a secondary model able to simulate the overall lethal effect of the treatment.

The aims of this study were (i) the selection of the most heat resistant L. monocytogenes among a pool of dairy origin strains, (ii) the estimation of D and z values of this strain using as matrices MBC cheese curd and the fluids coming into contact with MBC cheese after cheese molding, such as hardening water, brine and fresh covering liquid; (iii) the proposal of a secondary model able to estimate the cumulative effect of the stretching process on L. monocytogenes survival. Following this approach, we aimed to improve the knowledge about the influence of the matrix on heat resistance and to better understand the role that cheese curd stretching plays in L. monocytogenes inactivation.

A large variability of heat resistance among the L. monocytogenes strains tested was observed in this study and it has to be taken into account in future studies on L. monocytogenes survival during heat processes of food products. Our results in TSB medium agree with previously reported data (Doyle et al., 2001; Aryani et al., 2015b). A similar high heat resistance was found by Quintavalla and Campanini (1991) for a strain isolated from meat and tested in the same matrix. Curiously, the most heat resistant strain among the 20 tested by Aryani et al. (2015b), with the majority of them of food origin, was isolated from milk. Otherwise, some strains showed a very low heat resistance; the most heat sensitive strain was a reference strain (LMG 21264, isolated from human meningitis), whose heat resistance was about 11 times lower than that of Lm15 and similar to Lm28 that was about 10 times less heat resistant than Lm15. De Jesús and Whiting (2003) found a narrower distribution from 1.98 to 5.31 min of the D_60^°c values among 21 L. monocytogenes strains isolated from humans and animals, while the range for the same acid-adapted strains was between 0.8 and 4.59 min.

The influence of the matrix on the heat resistance of the selected strain used in the present study was clear and it was probably linked to a combination of protective and inhibiting factors. The high amount of total solids of the drained cheese curd (63.4%) and of the synthetic TSB medium (31.5 g/L) may be related to the high D-values measured at all the tested temperatures. No literature data are available about the heat resistance of L. monocytogenes in MBC cheese curds, while D_60^°c values in TSB (4.74 min corresponding to 284.4 s in Table 2) resulted similar to those reported by Quintavalla and Campanini (1991) and Aryani et al. (2015b). A higher thermotolerance of L. monocytogenes in half cream, double cream or butter compared to TSB was reported by Casadei et al. (1998), but no relationship was found between the D-values and the fat amount of the product. The presence of undissociated lactic acid in the covering liquid may be responsible for the lowest D-values reported in the present study. As the fresh covering liquid contained 0.5 g of lactic acid/100 mL at pH 2.9, according to the Henderson-Hasselbalch equation.

The amount of undissociated acid is near to 90% corresponding to about 50 mmol/L, largely above the concentration of 6.35 mM indicated by Aryani et al. (2015a) and Wemmenhove et al. (2018) as inhibitory of L. monocytogenes growth. Considering the values obtained in hardening water and brine, the relatively low NaCl concentration (4 g/100 g) of the latter fluid did not markedly affect Lm15 heat resistance, probably because the protective effect of salt was counterbalanced by the low pH value of the brine. The direct relation between salt concentration and L. monocytogenes thermotolerance was observed by several authors (Jørgensen et al., 1995; Juneja et al., 2013; Li et al., 2017).

The calculation of z-values between 60 and 70°C in the different matrices gave results generally higher (from 8.2 to 11.1°C) than most of the data present in literature, with the exception of z-value related to TSB medium (6.4°C). However, similar or higher z-values were found for L. monocytogenes strains tested in some foods such as egg albumen or yolk, meat or flour (Palumbo et al., 1996; Michalski et al., 2000; Taylor et al., 2018). A higher z-value means that the strain is more tolerant to the changes in temperatures and this behavior should be taken into account in the design of heat processes.

As the knowledge of D- and z-values is pivotal to successfully design food processes, the accuracy of the simulation of the lethal effect due to stretching and other heating operations depends both on the accuracy of the D- and z-value estimation, both on the accuracy of the determination of the heat treatment. While temperature changes during fluid heating by flowing through continuous heat exchangers can be predicted by computational fluid dynamics (Rinaldi et al., 2018) or direct measurements, the measurement of the change of temperature of a cheese curd during continuous cooking-stretching operation in industrial conditions, by means of the typical diving arms equipment used in the geographical area, is difficult to be monitored or estimated, and for this reason is lacking in literature. Cheese curd stretching is considered as an asymmetrical heat treatment (Villani et al., 1996; Murru et al., 2018) and just few known examples (Serraino et al., 2013) of description of the temperature evolution are referred to lab scale experiments or to industrial trials performed with batch operations (Mucchetti et al., 1997). The process involves solid/fluid mixing operations, rheological changes due to curd melting, and a flow in a channel with not constant section, responsible for a variable speed and residence time in the different sections of the equipment (tween screw conveyor, diving arms, and molding). The process can be divided in four steps: (i) a first step of fast rise of temperature of the curd mass, which homogeneity depends on the efficacy of cheese curd mixing with hot water into the horizontal counter rotating twin screw conveyor that transports the mixture to the kneading section; (ii) a second step where the melted cheese curd is kneaded by the movement of diving arms and temperature is quite constant and homogenous within the mass; (iii) a third step where the stretched cheese mass enters the molding equipment and it is pulled by a second screw conveyor toward the molding drum; (iv) a final step during which the molded cheese balls fall from the molding drum into flowing tap water where they start to harden and to decrease their temperature; the temperature decrease in the center of the cheese, that is slower than in its external part, is governed by the conductive heat transfer from the center to the surface of the cheese and it is dependent of the cheese ball size. The minimum overall sterilizing effect on L. monocytogenes is the result of the cumulative effect of non-isothermal (the steps of curd mixing with hot water and cheese chilling by dipping in tap water) and quasi-isothermal (the steps of curd kneading and cheese molding, performed without water addition) contributions. The accuracy of lethal effect prediction depends on the correct measurement of the temperature profile of the coldest spot in the non-isothermal steps of the process. This discussion shows the complexity of the phenomenon of heat exchange during the curd stretching operation and the difficulties to foresee the rate of inactivation of microorganisms during cheese curd stretching and cheese hardening. At the same time, beside the strain biodiversity (Murru et al., 2018), the potential process variability can contribute to explain the heterogeneous results obtained by challenge tests, where the inactivation of L. monocytogenes ranged from 1 to more than 8 decimal reductions (Villani et al., 1996; Kim et al., 1998; Raimundo et al., 2013; Serraino et al., 2013; Murru et al., 2018). The technological importance of bacterial heterogeneity is also connected to the ability of a small fraction of any population to survive exposure to stresses that kill the majority of the population. The knowledge of heat inactivation kinetics parameters for all the other fluids involved in chilling, salting and storing of MBC cheese is critical to more effectively and efficiently manage the treatment of these fluids to prevent MBC cheese surface post contamination.

The data supplied in this in vitro study may prove to be useful for MBC and other Mozzarella cheese producers in determining appropriate durations and temperatures for producing fresh pasta filata cheeses avoiding the presence of L. monocytogenes. To do so, a further model should be studied to foresee the amount of hot water required to heat the MBC cheese curd mass, considering both the cases of batch or in continuous stretching processes. Finally, the knowledge of the kinetic parameters of heat inactivation together with a deeper knowledge of the temperature vs time evolution of the cheese mass can be a tool to foresee the lethal effect of the process and to better manage the process itself also With the aim to improve the microbial safety of the cheese.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AR: investigation and methodology. MA: methodology and data analysis. FM: investigation and methodology. VB: project administration, supervision, and writing -review and editing. AG: resources. GP: resources. EN: supervision, writing -review and editing. GM: conceptualization, supervision, and writing – original draft. All authors contributed to the article and approved the submitted version.

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.