Introduction

Salt is already mentioned in the Old Testament as a food additive with a ritual character. Its use as a food preservative was known not only in Ancient Egypt and the Middle East but also in Ancient Rome [14]. Many foods are prepared containing common salt as a main ingredient and employed to flavour and preserve foods of various kinds, especially meat, fish and vegetables. Common salt, principally constituted by sodium chloride, lowers the water activity (a w) of a system [24] and thus renders conditions less favourable to microbial growth. Also, this additive can reduce the solubility of oxygen in water. Hence, the quantity of oxygen available to aerobic microorganisms in products containing high concentrations of sodium chloride is lower [14]. Many of the key systems used by the microorganisms to respond to either the loss or the gain of water due to changes in the osmolarity of the environment are now beginning to be understood, but the specific details of how each organism deals with these problems are different and are not completely known yet [21].

Increasing concern about hypertension in humans, which is correlated with cardiovascular diseases, has prompted food processors to develop low-sodium products. Sodium chloride is contraindicated for certain heart diseases, the circulatory system and the kidneys. The negative effects related to high sodium chloride concentration can be reduced by replacing NaCl with an equivalent amount of other salts like potassium or calcium chloride. Tsapatsaris and Kotzekidou [25] studied the effects of the substitution of NaCl by 50% KCl on Lactobacillus plantarum and Debaryomyces hansenii growth. Naewbanij et al. [19] also showed the influence of various chloride salts on L. plantarum growth in cucumber extract. However, a detailed study of the effects of different chloride salts and their combinations on microbial growth is not available.

In this context, predictive microbiology could be an efficient tool. The effects of environmental factors on growth parameters (μ max, λ, and N max) can be described by predictive models [16]. The coefficients of response surface (RS) on the growth parameters, fitted to the environmental factors, can give estimations of the changes induced by each variable [18]. In the area of food research, RS methodology has been extensively applied to the treatment of data obtained by physicochemical and sensory analysis or nutritional parameters [6] and, of course, to study microbial growth [2, 25]. The D-optimal design process for mixture designs works exactly the same way as for RS. The design space is defined by the low- and high-level constrains on each factor and any multifactor constraints. The D-optimal selection chooses points from the candidate point set that are spread throughout the design region. The points selected are model dependent and are selected to minimize the variance associated with the coefficients. Therefore, the design should be augmented to provide for estimates of pure error by replication and for determining the lack of fit using excess design points. The primary difference between a standard RS experiment and a mixture experiment is that a special type of design must be used and the form of the mixture polynomial is slightly different from those used in RS. The coordinate system for mixture proportions is a simple coordinate system. Each of the vertices corresponds to a pure blend and each of the sides represents a mixture of two components and none of the others. Interior points in the triangle are mixtures in which all ingredients are present. The centroid of the triangle corresponds to mixtures with equal proportions of all ingredients [18].

Saccharomyces cerevisiae is a microorganism with a general application food manufacturing. It is very important in many food fermentations such as bread, wine, beer and table olives, but it also causes spoilage by refermentation after packaging regarding table olives, bottling of wines, beverages, and fruit concentrates and preserves [10, 23]. López-Malo et al. [15] and Arroyo López et al. [3] studied the influence of water activity and NaCl concentrations on the growth–no growth interface of S. cerevisiae in laboratory media. Almagro et al. [1] compared the effects of Na+ and K+ on growth and thermal death of D. hansenii and S. cerevisiae under stress conditions, using individual concentrations of NaCl and KCl. However, there are no detailed studies on the influence of different chloride salts in the growth response of this microorganism.

The purpose of this work was to investigate the effects of different concentrations of sodium, potassium, calcium, and magnesium chloride salts on the growth parameters of S. cerevisiae, using a D-optimal mixture design with constrains, and quantifying such effects by polynomial equations. The information obtained by this means can be used to control vegetable fermentations and as hurdles for packed foods in which this yeast could be present.

Materials and methods

Origin and identification of the yeast specie

The yeast used in this study was S. cerevisiae (reference IGAL01), previously isolated from table olives fermentations. It was identified using API 20C AUX (bioMérieux, Marcy l’Etiole, France) together with the restriction pattern generated with endonucleases CfoI, HaeIII, HinfI and ScrFI (Roche Molecular Biochemicals, Germany) from the polymerase chain reaction (PCR) to amplified 5.8.S rRNA gene and the two internal transcribed spacers (ITS1 and ITS2) [8].

Inoculum preparation and growth medium

One single colony from pure cultures of S. cerevisiae was inoculated into 5 ml of yeast–malt–peptone–glucose broth medium (YM, Difco™, Becton and Dickinson Company, Sparks, USA). After 48 h of incubation at 30 °C, the tubes were centrifuged, and the pellet was washed with sterile peptone water (0.1%, w/v), centrifuged and re-suspended again in a sterile peptone water to obtain a concentration of 7.40 ± 0.20 log10 CFU/ml.

The basal growth medium selected for all experiments was YM broth, modified with different proportions of chloride salts (Table 1). The salts used were NaCl, KCl, CaCl2 and MgCl2 (assay of 99.9%) supplied by Panreac Química SA (Castellar del Vallés, Barcelona, Spain). After the addition of salts, 100-ml aliquots of medium were dispensed into 250-ml bottles and sterilized at 115 °C for 10 min. Values of pH after sterilization was ≈5.5. Subsequently, the bottles were inoculated with 10 μl of the suspension of S. cerevisiae, reaching an initial concentration of inoculum of 3.59 ± 0.03 log10 CFU/ml. Water activity (a w) was measured by triplicate in each medium after the sterilization with an Aqua Lab metre (Decagon Devices Inc., Pullmant, USA) at 25 °C. Average values resulting from the combination of salts in each run (±standard deviation) are shown in Table 1.

Table 1 Treatments included in the D-optimal experimental mixture design, using the diverse mixture of sodium, potassium, calcium and magnesium chloride salts with their respective water activities (a w)
Full size table

Experimental design

The plan was generated by Design-Expert v.6 software (Stat-Easy Inc., Minneapolis, USA), according to a D-optimal experimental mixture design with constraint (NaCl + KCl + CaCl2 + MgCl2 ≤ 9.0%, w/v), which, in general, aims at the minimization of uncertainty in the global parameter estimation. A detailed discussion of the D-optimal designs can be found in [18] and application of its use in [2]. The design consisted of 20 selected level combinations of the quantitative variables plus four replicates, chosen also by the software to provide estimation of the pure error; then, the total number of treatments was 24 (Table 1). The units habitually used at industrial scale (%, equivalent to g per 100 ml) have been intentionally chosen to facilitate the possible application of results directly to production. The experiments were carried out at 30 °C.

Samples were taken from the bottles of the experiments at predetermined time intervals, established according to the yeast growth conditions, and plated onto YM agar plates. The total number of samples ranged from 8 to 15, distributed throughout the different phases of yeast growth (lag, exponential and stationary) and the experiments were monitorized at least for 10 days. Cells were diluted in sterile peptone water and surface spread using a spiral plate maker model dwScientific (Don Whitley Scientific Limited, England). All the plates were incubated aerobically at 30 °C for 48 h and counted using a CounterMat v.3.10 (IUL, Barcelona, Spain) imagine analysis system.

Effects on the growth parameters

The growth data from plate counts were enumerated as log10 values. The model of [4] was fitted to the growth curves by using the MICROFIT 1.0 software (Institute of Food Research, Norwich, UK) and the growth parameters final population cell density (N max), maximum specific growth rate (μ max), and lag time (λ) estimated (primary model).

To describe the combined effect of environmental factors (chloride salts) on growth parameters, a multiple quadratic regression models was applied to the design and responses (secondary model). It is represented in the following equation, expressed in the canonical (Sheffé) form:

$$ P = \sum\limits_1^4 {\beta _i x_i } + \sum {\sum\limits_{i < j}^4 {\beta _{ij} x_i x_j } } + \sum\limits_1^4 {\beta _{ii} x_{ii}^2 } $$
(1)

where x 1, x 2, x 3, and x 4 stand for NaCl, KCl, CaCl2, and MgCl2, respectively, P is the growth parameter (μ max, λ, and N max) or its transformation, and the βs are the coefficients to be estimated. An analysis of variance was produced for each of the response variables, testing the value of the applied model and determining if a more complex model could have a better fit. If the F test for the model was significant (P < 0.05) there was evidence that the model could explain the variations in the responses. The backward option (alpha to exit = 0.05) was used to retain the significant coefficients. In the final model, coefficients required to support hierarchy were also maintained. This was necessary because only hierarchical models are scale-independent and can be translated into actual units. Statistical analysis was carried out using the value of each replicate, in order to test the lack of fit test. Fit adequacy was checked by different diagnostic tests, which included the presence of outliers, Cook’s distance, leverage, graphs of residual versus predicted values, and the graphic representation of predicted versus experimental values.

The terms in the canonical mixture polynomial have simple interpretations. Geometrically, the parameter β i represents the expected response to the pure mixture x i  = 1, x j  = 0, and is the height of the mixture surface at the vertex x i  = 1. Usually, the portion of each polynomial given by \( \sum\limits_1^q {\beta _i x_i } \)(q = number of variables) is called the linear blending portion. The two factor interaction \( \left( {\sum {\sum\limits_{i < j}^4 {\beta _{ij} x_i x_j } } } \right) \) represents the excess response over the linear one and is often called the synergism (or antagonism, depending on the sign) due to non-linear blending. Quadratic terms were not additionally significant while cubic terms were always aliased. The equations in terms of actual values of the variables included also the intercept.

The effect of each factor was further assessed by the use of the trace (perturbation) plot to show how the response changes as each factor moves from chosen reference points, with all other factors held constant at reference values [22]. The response model was defined as:

$$ \hat Y = f\left( {x_1 ,\,x_2 ,\,x_3 ,\,x_4 } \right) $$
(2)

with \( x_1^* ,\,x_2^* ,\,x_3^* ,\;{\text{and}}\;x_4^* \) at chosen specific values of the factors (0, 0, 0, 0) and (5, 0, 0, 0).

The perturbation effect of x 1 was defined as follows:

$$ \hat Y{\text{(}}x_1 {\text{)}} = f\left( {x_1 ,\,x_2^* ,\,x_3^* ,\,x_4^* } \right) $$
(3)

The perturbation effects of x 2, x 3 and x 4 were defined similarly.

Trace curves were produced with the vertical axis representing Ŷ(x i ), in actual values, and the horizontal axis representing x i (in coded values). By overlaying all the response trace plots, the joined trace plot was obtained [13].

Model validation

For the model validation, a new series of independent experiments were carried out. Levels were chosen at random within the range of the values used in the design. Validation treatments were: (1) 4% NaCl, 2% KCl, 2% CaCl2, 1% MgCl2; (2) 1% NaCl, 4% CaCl2; (3) 1% KCl, 1% MgCl2; (4) 5% NaCl, 1% KCl, 2% CaCl2; and (5) 1% NaCl, 1% CaCl2, 1% MgCl2. Salts not mentioned in the respective treatments were absent from them. Growth parameters of S. cerevisiae were directly estimated from these experiments, and compared to those predicted by the model using the accuracy (A) and bias (B) factors as described by Baranyi and Pin [5].

Results

Contribution of each salt and their mixture to the water activity of the media

The experimental design and responses, expressed as a w, μ max, λ, and N max, are shown in Tables 1 and 2. The highest a w was observed in the culture media (average 0.996) and the lowest (0.953) at 9.0% NaCl concentration (Table 1). The a w for 9.0% KCl, CaCl2, and MgCl2, were 0.969, 0.956, and 0.978, respectively. Then, the addition of CaCl2 showed an a w decrease close to that of NaCl, followed by KCl and MgCl2. The rest of the mixtures had aw that ranged between these extreme values.

Table 2 Growth parameters for S. cerevisiae obtained by applying the primary model of [4] to the growth curves in laboratory media containing the diverse mixture of sodium, potassium, calcium and magnesium chloride salts
Full size table

The contribution of each salt to the total a w was estimated by means of the polynomial equation adjusted to a w as a function of the NaCl, KCl, CaCl2 and MgCl2, concentrations. Only linear terms were significant. The model F value (42.884,4) gave a P < 0.0001, which implied that the model was significant. The lack of fit was insignificant (F 15,4 = 1.64, P = 0.3376).

The equation which related a w with the content of the diverse salts, in terms of the actual units (% each salt), was the following:

$$ a_{\text{w}} = 0.9973 - 0.00485{\text{NaCl}} - 0.00335{\text{KCl}} - 0.00385{\text{CaCl}}_2 - 0.00221{\text{MgCl}}_2 $$
(4)

The negative sign of NaCl, CaCl2, KCl, and MgCl2 in the equation represents that all of them contribute to decrease a w with respect to that obtained in the absence of salts. This equation permits the calculation of a w in a laboratory medium containing any combination of the studied salts within the experimental region (restricted to NaCl + KCl + CaCl2 + MgCl2 ≤ 9.0%). The equation may also be of guidance for estimating a w in brines containing mixtures of these salts.

As the equation only contains linear terms, the contribution of each salt to the a w is proportional to the corresponding coefficients in coded values. The highest contribution to an a w decrease was due to NaCl (A) (coefficient −0.022 for x 1 in coded values), followed by CaCl2 (C) (−0.017, x 3), KCl (B) (−0.015, x 2), and MgCl2 (D) (−0.0099, x 4), at the same concentration (%) of each salt.

When the effect was measured as the a w deviation from the reference point containing 5% NaCl (5% NaCl, 0% KCl, 0% CaCl2, and 0% MgCl2), there was always a further decrease of a w as the NaCl, CaCl2, KCl or MgCl2 were progressively added (Fig. 1, upper left). Then, any initial brine composed of NaCl may be complemented with any other salt with an additional a w decrease that will depend on the type of salt added, following the effect in the order: NaCl > CaCl2 > KCl > MgCl2.

Fig. 1
figure 1

Trace plot of a w, μ max (h−1), 1/√λ, and N max (log10 CFU/ml), in actual values, as a function of the NaCl (A), KCl (B), CaCl2 (C), and MgCl2 (D) concentrations (in coded values) for the (5, 0, 0, 0), reference points. The effects are expressed as deviations from the reference point

Full size image

Effect of each salt and their mixture on the growth parameters of S. cerevisiae

Effect on maximum specific growth rate (μ max)

The model suggested by the sequential sums of square analysis included linear and two-way interactions (Table 3). Applying the backward option to it, the chosen model had a non-significant lack of fit (P > 0.3351) and explained a high proportion of the total variance (R 2 = 0.8812) while retaining five coefficients plus those required to maintain hierarchy (essential to estimate the equation in actual values). The information related to the statistical analysis of the model and their selected terms, in coded terms, are shown in Table 4.

Table 3 Sequential model sums of squares for μ max, λ, N max
Full size table
Table 4 ANOVA of the model and coefficient estimates, using backward elimination, for the response factor Y a (μ max, h−1), using the diverse mixture of sodium, potassium, calcium and magnesium chloride salts
Full size table

The effect of each salt individually on μ max can be deduced from the slopes of the variables in the coded equation when all others were absent. When a moderate proportion of NaCl (A) was present (reference point 5%, 0, 0, 0), the addition of a further amount of NaCl (A) and CaCl2 (C) had a decreasing effect on μ max but increasing proportions of MgCl2 (D) and KCl (B) had a stimulating effect on μ max (Fig. 1, upper right).

The equation of μ max as a function of the actual values of the variables is convenient because it leads to a more intuitive interpretation of their terms. It was:

$$ \begin{gathered} \mu _{{ \max }} = 0.5696 - 0.0563{\text{NaCl}} + 0.00079{\text{KCl}} - 0.03588{\text{CaCl}}_2 - 0.00433{\text{MgCl}}_2 + 0.000966{\text{NaCl}} \times {\text{KCl}} + 0.01034{\text{NaCl}} \times {\text{MgCl}}_2 - 0.00673{\text{KCl}} \times {\text{CaCl}}_2 + 0.01222{\text{CaCl}}_2 \times {\text{MgCl}}_2 \hfill \\ \end{gathered} $$
(5)

The plots of the significant interactions in the final model (Fig. 2) can help for its interpretation. The presence of increasing amounts of NaCl or CaCl2 decreased μ max but the addition of either KCl (Fig. 2, left) or MgCl2 (Fig. 2, centre and right) always increased μ max particularly at low levels of NaCl and CaCl2. In both cases, μ max reached a plateau around 0.54–0.57 h−1. The addition of KCl to the solution already containing CaCl2 had only slight increasing effect onμ max (data not shown).

Fig. 2
figure 2

Surface and contour plots of μ max (h−1) as a function of the binary mixture of the significant effects of the equation showed in Table 4, expressed in terms of actual value concentrations. Concentrations of the salts not in each of the graphs were set a 0%

Full size image

Effect on lag phase duration (λ)

In the analysis of this response, the Box–Cox plot for power transformation yielded a lambda = −0.5, which suggested an inverse square root transformation for variance stabilization. The model suggested by the sequential sums of square analysis of the transformed data included linear and two-way interactions (Table 3). The ANOVA corresponding to the reduced model obtained after application of the backward option, details of the fit, the coefficients retained, their standard errors and probabilities as well as the complete equation in coded terms (at the bottom), are shown in Table 5.

Table 5 ANOVA of the model and coefficient estimates, using backward elimination, for the response factor Y a (1/√λ), using the diverse mixture of sodium, potassium, calcium and magnesium chloride salts
Full size table

The lag phase was mainly influenced by MgCl2 and NaCl and the interactions NaCl × CaCl2, CaCl2 × MgCl2, and KCl × CaCl2 but the interactions NaCl × KCl, Na × MgCl2, and KCl × MgCl2 were also significant. The linear effects of KCl and CaCl2 were removed by the reduction procedure but were introduced again in application of the hierarchical principle. The individual effects of other salts in presence of an initial 5% NaCl (reference point 5%, 0, 0, 0), obtained as already described for μ max, is shown in Fig. 1 (lower, left). The further addition of NaCl (A) and CaCl2 (C) increased λ, with a more pronounced slope for NaCl. The lower effect of CaCl2 than that due to NaCl when there is already a 5% NaCl can be due to the previous content of NaCl and the global constrain (NaCl + CaCl2 ≤ 9.0%), which limits the additional CaCl2 concentration that can be added and therefore its influence. KCl had a negligible effect and MgCl2 decrease λ.

The equation of the lag phase as a function of the diverse salt contents, in terms of the actual values, was the following:

$$ \begin{gathered} \frac{1}{{\sqrt {\lambda { }} }} = 0.449 - 0.031{\text{NaCl}} - 0.012{\text{KCl}} - 0.041{\text{CaCl}}_2 - 0.006{\text{MgCl}}_2 + 0.002{\text{NaCl}} \times {\text{KCl}} + 0.006{\text{NaCl}} \times {\text{CaCl}}_2 + 0.003{\text{NaCl}} \times {\text{MgCl}}_2 - 0.004{\text{KCl}} \times {\text{CaCl}}_2 + 0.002{\text{KCl}} \times {\text{MgCl}}_2 + 0.008{\text{CaCl}}_2 \times {\text{MgCl}}_2 \hfill \\ \end{gathered} $$
(6)

Interpretation of the equation, expressed in transformed response is difficult. Furthermore, the equation of 1/√λ cannot be visualized as a function of all the design’s variables. The effect of the most significant (P < 0.05) interaction can visualize the outstanding features (Fig. 3). The interaction NaCl × CaCl2 was marked and, for the same NaCl, there was always an increase in λ period as the CaCl2 increased. Addition of CaCl2 to one solution already containing KCl led always to an increase in λ but the incorporation of MgCl2 to solutions of CaCl2 decrease λ.

Fig. 3
figure 3

Contour plots of λ (h) as a function of the NaCl, KCl, and MgCl2 interactions with CaCl2 from the equation showed in Table 5. Concentrations and original response are expressed in terms of actual values. Concentrations of the salts not in the graph were set a 0%

Full size image

Effect on maximum population reached (N max)

N max in the diverse runs ranged from 6.46 to 7.64 log10 CFU/ml. Thus, the final populations were less influenced by the presence of the diverse salts than μ max and λ. The model suggested by the sequential sum of square analysis included linear and two-way interactions (Table 3). By applying the backward option to the full equation, it was possible to obtain a reduced significant model (Table 6), although with a non-significant lack of fit. A non-significant lack of fit would have been preferred, but the model chosen may be considered as a reasonable option in this case because the alternative model, special cubic, which showed a non-significant lack of fit included aliased variables and had negative predicted R 2. The coefficient retained (NaCl, KCl, and CaCl2 linear effects were included in application of the hierarchy principle), their standard errors, F values, probability and the equation in coded values are shown in Table 6.

Table 6 ANOVA of the model and coefficient estimates, using backward elimination, for the response factor Y a (N max, log10 CFU/ml), using the diverse mixture of sodium, potassium, calcium and magnesium chloride salts
Full size table

The effect of each other salt in presence of an initial 5% NaCl (reference point: 5% NaCl, 0, 0, 0) is shown also in Fig. 1 (lower right). The population size of S. cerevisiae could only be increased by the addition of MgCl2 while the incorporation of any of the other salts always decreased the population size; the effect was higher (lower N max) for CaCl2, followed by NaCl, and KCl, which induced a less marked N max decrease. The order of effects was not the same as when each salt was studied individually due, possibly, to the disturbance created by the NaCl content already in the solution.

The equation of the polynomial for N max in actual values was:

$$ \begin{gathered} N_{\max } = 7.64 - 0.097{\text{NaCl}} - 0.040{\text{KCl}} - 0.126{\text{CaCl}}_2 - 0.016{\text{MgCl}}_2 + 0.0162{\text{NaCl}} \times {\text{MgCl}}_2 + 0.015{\text{KCl}} \times {\text{MgCl}}_2 \hfill \\ + 0.031{\text{CaCl}}_2 \times {\text{MgCl}}_2 \hfill \\ \end{gathered} $$
(7)

The graphical study of the three significant interactions is shown in Fig. 4. It was always observed that the addition of MgCl2 salt to a solution already containing any other salt increased N max. High N max depended on the respective salt concentrations, but always required relatively low concentrations of NaCl (≈2.0%), KCl (≈2.25–4.60%), and CaCl2 (≈2.25%) and relatively high concentrations of MgCl2.

Fig. 4
figure 4

Contour plots of N max (log10 CFU/ml) as a function of the binary mixtures of the significant effects of the equation showed in Table 6, expressed in terms of actual values. Concentrations of the salts not in each of the graphs were set a 0%

Full size image

Validation results

The accuracy (A) and bias (B) factors, obtained for μ max, λ, and N max from independent experiments are shown in Table 7. The best results were obtained for N max (A and B ≈ 1), while the worst predictions were for the bias parameter of μ m (1.154) but, in general, values can be considered acceptable. The models obtained could then be used to navigate throughout the experimental region and make predictions with respect to the probable values for the growth parameters of S. cerevisiae in the presence of NaCl, KCl, CaCl2, and MgCl2 salt or their mixtures.

Table 7 Accuracy (A) and bias (B) factors [5] obtained for μ max, λ, and N max from independent experiments within experimental region (NaCl + KCl + CaCl2 + MgCl2 ≤ 9.0%)
Full size table

Discussion

The impact on growth of some spoilage bacteria and pathogenic organisms has been investigated in model systems of either broth or meat emulsions where Na+ from sodium chloride was substituted with other cations, i.e. K+, Mg++ and Ca++, but maintaining similar a w and pH. The strongest inhibition was observed with divalent cations [20] but the effect of replacing half or all of the Na+ content with K+ in meat slurries was either rather small or negligible [20]. The growth of Listeria innocua in broth and the surface of gelatine were controlled by a w in the range 0.91–0.99 and 0.92–0.99, respectively, by the addition of solutes such as NaCl, KCl, glucose, sucrose, and glycerol. The shortest generation time was observed in broth with NaCl followed by gelatine with NaCl, but removal of water was more stressful than the addition of NaCl [12].

NaCl in vegetable fermentations could also be substituted with other chloride salts, especially those with favourable effects on nutritional value of the final products. The most studied compounds have been KCl and CaCl2, but MgCl2 was also included in this research. Viander et al. [26] studied the impact of low NaCl, in combination with different proportions of KCl, on the spontaneous large-scale fermentation process of white cabbage into sauerkraut juice. The changes did not influence the pH decrease and acetic acid production but the lactic acid yield in the presence of KCl was clearly lower. The presence of calcium affected the microbial quality and consistency of chile pepper (Capsicum annum) mash during fermentation. The addition of a concentration at 8 or 15 g/100 g released more soluble sugars, increased alcohol production and reduced pH faster than the control or any other treatment. The CaCl2 was also effective in controlling the microbial growth [9]. The natural content of calcium in olives is fairly high [10] but its effect on a w of fermentation brines had not been studied yet.

The estimation of a w can be achieved by considering the solutes in solution (1975), or these and the non-solutes [7]. In this work, it was preferred to directly measure a w in the diverse treatments and then find the function which linked it with the salt concentrations. The relationship can be used as an approach to explore their effects on other fermented foods. The presence of the habitual NaCl concentration in the brines of fermented or stored olives (8–9%) produces only a limited decrease in the a w and a complete stabilization would require higher concentrations incompatible with the organoleptic and nutritional characteristics [10]. The antimicrobial properties of NaCl alone, used for the preservation of natural casings were sufficient to reduce the bacterial contamination below acceptable levels (except for Clostridium spores) at a w below 0.850 [27]. Other chloride salts could have similar trend. The effect of supplementing a moderate level of NaCl (5%) with CaCl2 caused a further decrease of μ max but the incorporation of KCl or MgCl2 stimulated growth. Furthermore, MgCl2 stimulated μ max in the presence of either NaCl, or CaCl2. As a result, the use of magnesium chloride, in addition to NaCl, could favour the growth of S. cerevisiae and, in consequence, the fermentation or spoilage rate, depending on the processing step. However, this yeast usually coexists with lactic acid bacteria and other micro-organisms and the behaviour in real fermentations cannot be extrapolated straightforward. Yeasts are a part of the natural micro flora of olive fermentation but the effect of a higher population of S. cerevisiae on physicochemical and organoleptic characteristics is not yet known [10]. As the presence of KCl and CaCl2 in olive fermentations is fairly high, some of the empirical practices to elaborate this product at industrial scale may be based on the above-mentioned relationships [10].

According to the results found in this work, the addition of NaCl and CaCl2 to a brine already containing NaCl, increased the λ period. Therefore, the addition of both salts could eventually be used alone or in combination to prevent a rapid development of S. cerevisiae not only in laboratory media but also in foods in general. This practice could improve stabilization of some foods like olives. The presence of CaCl2 may not only lead to a longer lag phase but may also be used to control S. cerevisiae final population due to its effect decreasing the N max. On the contrary, addition of MgCl2 can stimulate the growth and N max in brines already containing NaCl or other salts.

Mulé et al. [17], used different types of brines when processing green table olives (lavorazione “al naturale”) to obtain a product with lower sodium content, suitable for those who suffer from hypertension, by partially substituting Na+ with Ca++ and K+ in the fermentation brines.

Regarding pickles on the effects of NaCl, KCl, CaCl2, and MaCl2 and their combinations could be useful. Calcium chloride and potassium sorbate (KS) reduced the NaCl used during natural cucumber fermentation and storage but the presence of CaCl2 helped to maintain cucumber firmness. A synergistic action between NaCl, CaCl2, and KS allowed good quality pickles using moderate amounts (5% NaCl, 0.2%CaCl2, and 0.2% KS) of the three components [11]. The effect of including KCl, CaCl2, and MgCl2 as partial substitutes for NaCl in the fermentation in cucumber extracts showed that the salts that gave the highest percentage of total lactic population, provided the most selective environment, were NaCl and KCl in the filtered, raw cucumber extract and KCl in the heated and filtered extract, while CaCl2 was more selective than MgCl2 in either medium [19]. Results from this work open the possibility to use diverse salt combinations.

Conclusions

Results from this work showed that the effects of similar proportions of diverse salts in laboratory media on a w and on the growth parameters of S. cerevisiae were diverse and could be quantified by polynomial equations as a function of their concentrations. The order of contribution to a w decrease was NaCl > CaCl2 > KCl > MgCl2. When the salts were added to a solution of NaCl (5%), further additions of any other salt followed a similar trend. The contribution of each salt to the μ max decrease followed the similar trend for NaCl and CaCl2; but, in the presence of a 5% NaCl solution, a further addition of KCl and MgCl2 markedly increased μ max. MgCl2 had a stimulating effect on μ max when used in combination with NaCl or CaCl2, while KCl followed the same trend in the presence of NaCl. The lag phase of S. cerevisiae was mainly affected by MgCl2 and the interactions NaCl × CaCl2 and CaCl2 × MgCl2. The further adding of NaCl and CaCl2 to a solution already containing 5% NaCl increased the length of the lag phase but KCl had negligible effect and MgCl2 slightly decreased it. The most outstanding effects on N max were due to MgCl2 and the interactions of MgCl2 with NaCl, KCl and CaCl2. When the salts were added to an initial solution already containing 5% NaCl, CaCl2 (strongly), NaCl and KCl decreased the N max but MgCl2 increased it; the incorporation of MgCl2 to NaCl, CaCl2 or KCl produced a stimulating effect on N max. These results can be applied to fermentation, storage and preservation of pickles or to any other fermented food in which S. cerevisiae is present.