The Joint Effect of pH Gradient and Glucose Feeding on the Growth Kinetics of Lactococcus lactis CECT 539 in Glucose-Limited Fed-Batch Cultures

In recent years, there has been an increasing interest in using probiotic cultures from lactic acid bacteria (LAB) as additives in animal feed to prevent or treat diseases (Alvarez-Olmos and Oberhelman 2001) and enhance the production results, e.g. weight gain and feed conversion efficiency (Guerra et al. 2007a).

For a successful application in animal feed, the probiotic cultures should contain a high concentration of viable cells, bacteriocins and fermentation metabolites to control the growth of pathogenic bacteria in both the animal feed and gut of the animals (Guerra et al. 2007a; Costas et al. 2018). Three alternatives for this purpose are: i) the use of cheaper fermentation and feeding media (like wastes from the food industry), ii) the selection of an appropriate strain, and iii) the design of an adequate fermentation procedure.

Whey (a waste from the cheese-making process) and mussel processing wastes (MPW) contain lactose in the relatively high concentration (~ 50 g/l in case of whey), glycogen (~ 10 g/l in case of MPW), proteins (~ 5.0 g/l in case of whey and 3.5 g/l in case of MPW), as well as micronutrients, including amino acids, vitamins and minerals (Murado et al. 1994; Costas et al. 2018). For these reasons, both substrates have been used for productions of probiotic biomass (Costas et al. 2018) and bacteriocins (Garsa et al. 2014; Costas et al. 2018) by different LAB.

The comparison of the antibacterial activity of 38 bacteriocin-producing LAB (including Lactococcus, Pediococcus, Lactobacillus, and Leuconostoc strains) against a set of 19 indicator bacteria (including Listeria, Enterococcus, Carnobacterium, Lactococcus, Lactobacillus, and Leuconostoc strains) showed that the nisin produced by L. lactis CECT 539 was the most potent bacteriocin with the broadest antibacterial spectrum (Guerra and Pastrana 2002). In addition, previous studies showed that the strain CECT 539 has potential as a probiotic additive for animal feed (Guerra et al. 2007a), or as a food products’ preservative to extend their shelf life (Guerra et al. 2005).

Considering that fed-batch fermentation allows obtaining cultures with a high concentration of viable cells (Cho et al. 2010; Costas et al. 2018), the use of this fermentation technique for production of highly concentrated probiotic cultures on whey and MPW could be very advantageous. With this approach, the high mean chemical oxygen demand of these wastes, about 70 g/l in case of whey (Slavov 2017) and 25 g/l in case of MPW (Murado et al. 1994), and consequently their contamination effects could be considerably reduced.

Previous studies showed that the growth and nisin synthesis by LAB in realkalized fed-batch cultures depend on the stepwise pH profiles generated in the cultures, due to i) the effect of pH on nutrient uptake, ii) the inhibitory effect of low pH values on biomass and product formation, and iii) the specific effect of pH on bacteriocin synthesis (Cabo et al. 2001; Costas et al. 2018). Then, the use of an adequate mathematical model describing the joint effect of the culture pH and nutrient (glucose) addition on the productions of biomass, nisin and fermentation products could provide a better understanding for controlling and optimizing the fermentation process.

However, to our knowledge, no information is available on the quantification of the joint effect of glucose addition with the feeding media and the pH gradients (VpH) generated in the cultures on the synthesis of biomass, nisin, lactic acid, acetic acid and butane-2,3-diol by L. lactis strains.

For these reasons, in the present study, a first glucose-limited realkalized fed-batch culture in DW medium was designed by using a medium prepared with mussel processing wastes (CMPW) and a 400 g/l concentrated lactose (CL) as feeding media. From the results obtained in this culture, the effects of glucose addition and VpH on biomass and product synthesis by L. lactis CECT 539 were quantified, for the first time, by using empirical modeling. To validate the effectiveness of this modeling procedure, a second glucose-limited realkalized fed-batch culture in DW medium was carried out by using the CMPW medium supplemented with glucose and KH₂PO₄ up to concentrations of 400 g glucose/l and 3.21 g total phosphorus/l as feeding medium.

Growth kinetics of L. lactis CECT 539 in the realkalized fed-batch cultures. The evolution of the culture variables (pH and concentrations of nutrients: glucose, lactose, proteins, total nitrogen, and phosphorus) in the cultures I and II are shown in Figures 1 and 2, respectively. Although the pH gradients decreased progressively in both cultures, L. lactis CECT 539 did not lose completely its capacity to recover the acidic pH at the end of fermentation. In both cultures, glucose was completely consumed in every realkalization and feeding cycle, but relatively high concentrations of lactose remained in the media (Figs. 1 and 2). The proteins and total nitrogen and phosphorus were consumed in parallel with the production of biomass, but the complete exhaustion of these nutrients was not observed in the cultures (Figs. 1, 2 and 3).

Fig. 1.

Fig. 2.

Fig. 3.

The nisin-producing strain was grown in two growth cycles (Fig. 3) composed of two exponential growth phases, separated by an intermediate lag phase, as observed before in previously realkalized fed-batch cultures in DW medium (Guerra et al. 2007b; Costas et al. 2016, 2018).

Since L. lactis CECT 539 exhibited a diauxic growth in a batch culture in DW medium supplemented with glucose (Costas et al. 2016), it could be logical to suppose that the nisin-producing strain presented two growth phases within every realkalization and feeding cycle of 12 h in the realkalized fed-batch cultures I and II. However, the diauxic growth pattern observed in both cultures would not be related with the use of glucose- and lactose-containing substrates (culture I) or with a glucose-containing substrate (culture II), as feeding media. This is mainly because the feeding media were added at the beginning of every realkalization and feeding cycle to bring the culture up to the initial total sugars concentration (22.62 g/l) of the fermentation (DW) medium (Guerra et al. 2007b; Costas et al. 2016, 2018), and the experimental X concentrations were obtained at the end of every cycle (Figs. 1, 2 and 3).

The TN and TP concentrations decreased progressively in the first culture as a consequence of their consumption by the nisin-producing strain and the sample collection every 12 h. However, in the second fermentation, the growth slowed down at 144 h of incubation, even though the culture was fed with a substrate (CMPWGP) containing a relatively high TP concentration (3.21 g/l) that produced an increase in the concentration of this nutrient in the fermentation medium (Fig. 2).

Thus, culture II was stopped at 168 h because the cells entered in the second stationary phase of growth after 144 h of incubation and, taking into account the profiles described by the variables X and Nis in fermentation I (left side of Fig. 3), it is logical to consider that the extension of the fermentation would not produce significant increases in the concentration of both variables.

In addition, the TN source could be considered as a limiting substrate for the growth of L. lactis CECT 539 in fermentations I and II, rather than the two carbon (glucose and lactose) or TP sources. Thus, the nisin-producing strain probably grown first consuming the free amino acids and utilizable oligopeptides (Letort et al. 2002) present in the DW medium and/or the inorganic nitrogen (Guerra et al. 2007b) added with the feeding media (CMPW or CMPWGP). In the second exponential growth phase, the strain CECT 539 probably utilized the less preferred nitrogen sources, for example, the proteins present in both the DW and CMPW media (Murado et al. 1994; Letort et al. 2002).

Nisin production paralleled both the biomass production and pH gradients generated in the two cultures (Figs. 1, 2 and 3), because this metabolite was produced as a pH-dependent primary product in this kind of realkalized fed-batch fermentations (Cabo et al. 2001; Guerra et al. 2007b; Costas et al. 2016; 2018). Lactic acid was also synthesized in parallel with biomass synthesis in both cultures, but the productions of acetic acid and butane-2,3-diol triggered after 84 h of fermentation (left and right sides of Fig. 3).

Although the incubation times in fermentation I (264 h) and fermentation II (168 h) were different, the final levels of X, VCC, Nis and AA synthesized in the first culture (3.07 g/l, 1.75 × 10¹⁰ CFU/ml, 105.61 BU/ml, and 1.78 g/l, respectively) were almost similar (p > 0.05) to those (3.08 g/l, 1.75 × 10¹⁰ CFU/ml, 103.68 BU/ml, and 1.66 g/l, respectively) obtained in fermentation II. Only the LA and B concentrations in the first culture (16.98 and 4.96 g/l) were higher (p < 0.05) than those (13.62 and 2.24 g/l) obtained in fermentation II (Fig. 3).

This observation indicates that the CMPWGP medium can be used as an appropriate feeding substrate for the production of probiotic biomass and nisin. However, further studies based on optimizing its TP concentration are required to avoid the accumulation of this nutrient in the fermentation medium.

Empirical models obtained for biomass and product synthesis in the first fed-batch culture. Table II shows the statistically significant values obtained for the constants of model [7] for the dependent variables (X, VCC, Nis, LA, AA and B) in fermentation I. In all cases, RPDM values < 6 % and R² values > 0.994 were obtained, indicating that the fitted models described appropriately the trend observed for all the dependent variables in response to the addition of glucose and the pH gradients generated in the culture. The fitted models obtained for each dependent variable are discussed below:

Biomass and viable cell counts. For both dependent variables (X and VCC), the values of the coefficients for the lineal (VpH and G) and quadratic (VpH² and G2) terms had a negative and positive sign, respectively (Table II). However, while the coefficient of the binary interaction between VpH and G had a positive sign for X, it was not statistically significant for VCC.

From a mathematical point of view, these results indicate that X and VCC increased for VpH and G values lower and higher than the corresponding optimum values (VpHopt and Gopt) for both independent variables. However, from the results obtained in previous works (Cabo et al. 2001; Guerra et al. 2007b), it could be reasonable to consider the obtained optimum VpH values (1.81 for X and 1.92 for VCC) as the minimum pH gradient that allows the growth of L. lactis CECT 539 in the first culture. In fact, the corresponding maximum final pH values, 5.19 for X and 5.08 for VCC (pHopt = 7.0 – VpHopt), were slightly higher than that (pH = 4.86) observed in the previous batch cultures of the nisin-producing strain in DW medium (Guerra et al. 2007b). These maximum final pH values for the growth of L. lactis CECT 539 are lower than that (5.50) observed for L. lactis ATCC 11454 in MRS broth (Penna and Moraes 2002).

The decrease in the bacterial growth at pH values lower than 5.19 or 5.08 could be related to a reduction in the metabolic activity of L. lactis CECT 539 probably caused by a limitation in micronutrients or nutrient transport (Poolman and Konings 1988).

Using the same argument as that used for the VpHopt, it can be also reasonable to consider the values of Gopt (1.59 g/l for X and 1.63 g/l for VCC) as the minimum concentrations of glucose that should be in the culture medium to stimulate the growth of L. lactis CECT 539.

Nisin. The statistically significant coefficient for the quadratic term VpH² with a negative sign indicates that the optimum VpH value (2.32) for maximum nisin production (Table II) is within the pH gradient range (from 2.80 to 1.33) generated in the first culture (Fig. 1). Thus, the calculated minimum final pH value (4.68) is in perfect agreement with the optimum final pH values (4.90 and 4.27) obtained for nisin synthesis in DW medium in batch and realkalized fed-batch cultures, respectively (Guerra et al. 2007b).

With regard to the addition of glucose, it can be noted that the Gopt value obtained (1.29 g/l) is the minimum value that favored bacteriocin production (Table II). Then, more addition of glucose in every realkalization and feeding cycle is needed to enhance nisin synthesis in fermentation I.

Lactic acid, acetic acid, and butane-2,3-diol. The signs of the lineal and quadratic terms in the empirical models obtained for these three metabolites were similar to those obtained for the dependent variables X and VCC. Thus, according to the predictions of model [7], the maximum final pH values for LA, AA and B productions were 4.70, 4.17 and 4.60, respectively.

Lactic acid production in the realkalized fed-batch culture I was minimal at final pH values between 4.62 and 4.83 (Fig. 2), which includes the maximum pH value (4.70) calculated from model [7]. Similarly, when the culture reached final pH values between 4.22 and 4.97 (12–84 h of incubation), the lowest AA and B productions were obtained, because, at these incubation times, the two fermentation metabolites had not been detected in the culture medium (Fig. 2). In fact, AA and B productions started when the culture reached a final pH value of 4.99 after 84 h of incubation (left side of Fig. 3).

In the same way, the minimum concentrations of glucose in the processes for high LA, AA and B production were 1.76, 1.41 and 2.80 g/l, respectively (Table II).

Validation of model [7]. The effectiveness of model [7] to describe the productions of biomass, bacteriocin and fermentation metabolites by L. lactis CECT 539 was validated by using the experimental data obtained from the second fed-batch culture. As can be observed, the empirical model [7] satisfactorily described the evolution of the dependent variables X, VCC, Nis, LA, AA and B (right side of Fig. 3), providing RPDM values < 4% and R2 values > 0.979 (Table III). In addition, the maximum final pH values were similar (for X, VCC, LA and B) or almost similar (for AA) to the corresponding calculated values in the fed-batch culture I (see Tables I and II).

However, in the case of nisin synthesis, the optimum pH value calculated (4.88) could be considered as the maximum final pH value for high bacteriocin production in every realkalization and feeding cycle, since the value of the coefficient for VpH² had a positive sign. This indicates that nisin synthesis is favored in this type of fed-batch cultures in the pH range of 4.68 and 4.88, which is within the optimum pH range observed in previous studies for nisin production by L. lactis CECT 539 (Guerra et al. 2007b; Costas et al. 2016, 2018).

As shown in Table IV, the optimum final pH values for nisin production by other Lactococcus strains (e.g. L. lactis ATCC 11454 and ATCC 7962) were not always within the optimum final pH range calculated for L. lactis CECT 539. In addition, other lactic acid bacteria, including the genera Leuconostoc, Pediococcus, Lactobacillus, and Carnobacterium produced the highest bacteriocin concentrations at final pH levels different to those observed for L. lactis strains. These results indicate that the effect of the final pH on the bacteriocin production depended on both the strain and culture medium used.

This different effect of the final pH values on the bacteriocin synthesis has been related with the need of an appropriate final pH range for the post-translational conversion of prebacteriocin to active bacteriocin (Yang and Ray 1994).

With regard to the glucose addition, it could be noted that with the exception of LA, the signs for the coefficients of G and G² terms in the models for the other dependent variables, had a positive and negative sign, respectively (Table III). Therefore, the calculated Gopt values are the maximum added glucose levels to obtain the maximum concentrations of X, VCC, Nis, AA and B. Therefore, with the use of a feeding medium (CMPWGP) with a concentration of glucose (400 g/l) higher than the CMPW medium (101.72 g/l), the amounts of glucose added to the growing culture in fermentation II were considerably higher than those added in fermentation I (Figs. 1 and 2). Thus, glucose levels higher than the calculated Gopt value in the second culture could produce substrate inhibition on the growth and product (Nis, AA and B) formation (Pongtharangkul and Demirci 2006; Costas et al. 2016, 2018) by L. lactis CECT 539. In addition, the inhibitory effect produced by higher levels of glucose on nisin production could be also related with the regulation that the carbon source produced on the synthesis or activity of prenisin-modifying enzymes (De Vuyst 1995).

These observations are consistent with the results obtained for the fed-batch production of nisin by L. lactis ATCC 11454 in a complex medium, with sucrose feeding at rates of 6, 7, 8, and 10 g/l/h (Lv et al. 2005). In this study, the growth was slightly affected by the different feeding rates, but the bacteriocin synthesis was different in the four cultures. Thus, the optimal feeding rate for nisin production was 7 g/l/h, but the addition of sucrose at feeding rates of 8 and 10 g/l/h led to the accumulation of the carbon source in the culture medium during feeding that inhibited the bacteriocin synthesis. For a feeding rate of 6 g/l/h, the sucrose was almost completely consumed by the nisin-producing strain and its remaining level in the culture medium was close to zero in the period of feeding. This low sucrose availability limited the growth and bacteriocin production (Lv et al. 2005).

The results obtained with the use of model [7] are in perfect agreement with the affirmation that nisin is produced as a pH-dependent primary metabolite since its production depends on both the biomass synthesis and the final culture pH in the medium (Yang and Ray 1994; Guerra et al. 2007b). Thus, biomass production by L. lactis CECT 539 was favored at culture pH values above 5.0, but a final pH range of 4.68–4.88 was needed for high nisin synthesis.

From a practical point of view, the modeling procedure used in this work could allow: i) determining the optimum pH and glucose ranges to obtain high levels of biomass, nisin and fermentation metabolites in realkalized fed-batch cultures, ii) providing an accurate interpretation of the fermentation kinetics taking into account the effects of the amount of glucose added and the pH gradient generated in every realkalization and feeding cycle on the growth and product synthesis by L. lactis CECT 539 and, iii) design feeding strategies to produce highly concentrated probiotic products with high concentrations of viable cells and nisin.