Introduction

Probiotics are recognized for enhancing growth performance and overall health in pigs, thereby contributing to safer, more sustainable food production. By modulating the gut microbiome, probiotics improve nutrient absorption and immune function, supporting better feed efficiency, growth rates, and overall health in livestock1. This then reduces the need for antibiotics, aligning with the push toward antibiotic-free animal farming and reducing the prevalence of pathogenic bacteria. Consequently, the demand for probiotics in pig feed is rising, necessitating efficient and economical industrial-scale production to meet modern livestock farming requirements. Such production of probiotics involves several key development steps, including strain selection, identification of an optimal fermentation medium, and downstream processing. When selecting probiotic strains for industrial production, it is essential to choose those with high yield potential, robustness, genetic stability, scalability, viability, safety, functional benefits, and cost-effectiveness to ensure efficient large-scale production while maintaining quality and compliance2. The probiotic strain used in this study was Pediococcus acidilactici strain 72 N (P72N) is a lactic acid bacterium (LAB) which was isolated from Thai indigenous pig feces, and which is known for its probiotic potential and safety, including the absence of antimicrobial resistance genes, as per European Food Safety Authority (EFSA) standards3. P72N has demonstrated antibacterial, antiviral, anti-conjugation, and anti-biofilm activities in vitro4,5,6. A double-coated microencapsulation protocol for P72N has also been developed, and its use has shown benefits for pig intestinal health and growth7,8,9.

Another critical factor in production efficiency is the fermentation medium, which can account for 30–40% of production costs10. Standard media like MRS and M17 are costly and may not provide optimal conditions for all LAB strains11. Therefore, modified media formulations and optimization tailored to the nutritional needs of specific strains have become an area of increased research. Optimization techniques like the One-Variable-at-A-Time (OVAT) method, Plackett-Burman design (PBD), response surface methodology (RSM), and factorial design help to optimize media composition for cost-effective production12. In industrial fermentation processes, carbon and nitrogen sources are vital in LAB fermentation. Adjusting the carbon source ratios based on the specific requirements of the microbial strains used is essential to prevent the presence of residual sugars in the fermentation medium. Nitrogen sources that come from food processing industries, such as whey protein, yeast extract and meat extract, have emerged as promising alternatives in growth media formulation10,11,13,14. Optimizing fermentation conditions such as temperature, pH values, and agitation speed is also crucial for maximizing yield and maintaining product quality in industrial fermentation processes12.

Studies have shown that growth media composition can significantly impact the expression of key probiotic traits, including resistance to bile salts, tolerance to low pH, and bioactive compound production15,16,17,18. Therefore, it is essential to assess the quality and efficacy of probiotics in newly modified media to ensure maintenance of these critical functional traits. LC-MS-based metabolomic analysis is an important tool for detecting and characterizing metabolic alterations in response to changes in growth media, allowing for a comprehensive understanding of cellular adaptations and metabolic shifts.

To date, most media optimization studies have prioritized cost-effectiveness and biomass yield, often neglecting the preservation of strain-specific probiotic traits. In contrast, this study presents a novel approach by developing a food-grade, cost-effective medium tailored specifically for the industrial-scale production of the Thai native swine-derived probiotic strain Pediococcus acidilactici 72 N (P72N), while simultaneously evaluating its functional integrity. By integrating growth performance analysis with detailed probiotic trait assessment and LC-MS-based metabolomic profiling, this work provides a comprehensive framework for medium optimization that ensures both economic feasibility and the maintenance of probiotic efficacy—an area largely underexplored in the current literature.

Results

One-Variable-at-A-Time (OVAT) method

The preliminary screening using the OVAT method aimed to select carbon source significantly impacting P72N cell production. The effects of carbon sources on the viable cell production of P72N at 24 h of incubation are shown in Fig. 1. P27N viable cell numbers in 10 g/L, 20/L, and 30 g/L concentrations of glucose were 7.92 ± 0.02, 8.02 ± 0.02, and 8.05 ± 0.02 log CFU/ml; for sucrose they were 7.89 ± 0.08, 7.66 ± 0.04, and 7.77 ± 0.07 log CFU/ml; and for dextrose they were 8.00 ± 0.03, 8.19 ± 0.06, and 7.86 ± 0.06 log CFU/ml (Supplementary Table S1). Since the highest cell production occurred in 20 g/L dextrose (p < 0.05), it was selected as the optimal carbon source for further experiments (Fig. 1).

Fig. 1
figure 1

The effect of carbon sources on the viable cell production of P72N at 24 h of incubation.

Plackett-Burman design (PBD)

The experimental matrix of Plackett-Burman design and the response (cell counts) for P72N after 24 h incubation are shown in Table 1, where viable cell production varied from 8.70 ± 0.10 to 8.99 ± 0.12 log CFU/ml. The standardized effects of the variables for P72N are illustrated as single columns in the Pareto charts shown in Fig. 2A. The column passing through the vertical (dotted) line indicates statistical significance, meaning that the variable had significant impact on cell count. Yeast extract was the only nitrogen source to have a large effect on viable cell production, and was the only nitrogen source having a significant effect (p < 0.0001) (Figure. 2A and Supplementary Table S2). The fitness of the experimental model can be explained by the value of the coefficient of determination (R2. The coefficient of determination was high (R2 = 0.9459), and the values of adjusted co-efficient of determination (adj-R2 = 0.9256) and predicted coefficient of determination (pred-R2 = 0.8783) were in reasonable agreement. The accuracy of the model was also demonstrated by a statistically insignificant lack of fit (p > 0.05). The non-significant lack of fit value in this experiment (p = 0.066) confirmed the adequacy of the model for describing the data (Table S2). Therefore, yeast extract was selected as the only nitrogen source for growth of P. acidilactici 72 N in subsequent RSM experimentation.

Table 1 Experimental matrix of the Plackett–Burman design and the response (cell count) of P72N after 24 h incubation. X1 beef extract; X2 yeast extract; X3 sweet Whey. (+ 1), highest concentration of variable; (− 1), lowest concentration of variable. Data are presented as the mean ± standard deviation of three replicates.
Fig. 2
figure 2

Pareto chart showing the standardized effects of three nitrogen sources on viable cell production of P72N. Grey and black columns represent positive and negative effects on viable cell production (A). Contour plot for viable cell production of P72N showing the interaction between dextrose and yeast extract (B).

Response surface methodology (RSM)

The RSM with central composite design (CCD) consisting of 13 runs, each with three replicates, and the cell numbers achieved after 24 h of incubation are shown in Table 2. The viable cell production varied from 8.88 ± 0.16 to 9.20 ± 0.06 log CFU/ml. Although the highest viable cell number (9.20 ± 0.06 log CFU/ml) was found in run 6, it was not significantly different from the cell numbers in experimental runs 1, 4, 5, 7, 10, 11, 12 and 13. The response cell numbers were analysed by applying multiple regression analysis. The second-order polynomial equation, which expressed the relationship between the predicted response and variables, was as follows:

Table 2 Experimental matrix of the central composite design (CCD) and response (cell count) of P72N after 24 h incubation. X1 dextrose; X2 yeast extract. (+ 1), highest concentration of variable; (− 1), lowest concentration of variable; (0), central concentration of variable; (-α, α), axial concentrations of variables. Data represent the mean ± standard deviations of three replicates. a−cValues with different superscripts are significantly different (p < 0.05).
$${\rm{Y}} = {\rm{7.9955 }} + {\rm{0.2827}{X_1}} + {\rm{0.2912}{X_2}} - {\rm{0.04877}{X_1}^2} - {\rm{0.02191}{X_2}^2} + {\rm{0.00961}{X_1}{X_2}}$$
(1)

where X1 and X2 are dextrose and yeast extract respectively.

The significance of the quadratic regression model with linear, squared, and interaction terms are shown in an ANOVA table (Supplementary Table S3). The determination coefficient (R2 was 0.9899, implying that 98.99% of the variation in the response was explained by the model. The values of adj-R2 (0.9827) and pred-R2 (0.9484) were in acceptable agreement. Furthermore, the lack of fit test (p = 0.212) verified that the model was suitable for explaining the data. The linear and quadratic effects of dextrose (X1) and yeast extract (X2) significantly affected cell numbers (p < 0.0001). Dextrose and yeast extract could act as limiting nutrients because the squared effects of the two variables were highly significant (p < 0.0001). The contour plot is presented as Fig. 3, illustrating the interactions between the two variables and deriving the optimal concentration for maximal cell production. Viable cell numbers increased at a dextrose concentration between 18 g/L to 28 g/L and at a yeast extract concentration between 55 g/L to 65 g/L, and declined at concentrations beyond that range. There was no interaction between variables (X1 × 2) because the contour plot was round, and the probability value was low (p > 0.05) (Fig. 2B and Table S3).

Fig. 3
figure 3

Fermentation of P72N in a 250mL flask using modified medium MM2 (A) and in a 500mL flask using MM2 compared to MRS media (B) at 37 °C for 24 h with an agitation speed of 120 rpm, showing cell numbers, pH values and residual sugars.

The regression model predicted that the optimal formulation contained 23 g/L dextrose and 61.5 g/L yeast extract, with a maximum biomass concentration of 9.21 log CFU/ml, with a 95% confidence interval ranging between 9.19 and 9.23 log CFU/ml. The predicted formulation was confirmed under the optimized conditions by performing independent experiments in triplicate, and the biomass was found to be 9.24 ± 0.03 log CFU/ml. The model showed good agreement with the predicted value, and the experimental value was within the 95% confidence interval range. However, when compared to the predicted formulation, the experimental formulation with a concentration of 20 g/L dextrose and 45 g/L yeast extract (costing US$2.42/L) was considered financially acceptable, whilst producing a viable cell number of 9.13 ± 0.02–9.16 ± 0.01 log CFU/ml. Thus, the formulation with a concentration of 20 g/L dextrose, 45 g/L yeast extract, 1 g/L Tween 80, 2 g/L C6H14N2O7, 2 g/L K2HPO4, 5 g/L CH3COONa, 0.1 g/L MgSO4.7H2O, and 0.05 g/L MnSO4.H2O was initially chosen as a modified media (MM1) for the growth of P72N.

Before proceeding with scale-up fermentation, the residual sugar content in MM1 was evaluated by fermenting P72N in a 250 ml flask for 24 h at 37 °C. A considerable amount of residual sugars was found in the medium at the end of the fermentation process. Therefore, the concentration of dextrose in the MM1 was reduced from 20 g to 10 g. This modification did not have a negative impact on viable cell production (Supplementary Figure S1). Consequently, a final formulation containing 10 g/L dextrose, 45 g/L yeast extract, 1 g/L Tween 80, 2 g/L C6H14N2O7, 2 g/L K2HPO4, 5 g/L CH3COONa, 0.1 g/L MgSO4.7H2O, and 0.05 g/L MnSO4.H2O was used as an alternative modified medium (MM2) for P72N in subsequent fermentation processes.

Fermentation in a 250mL flask

The cell numbers, pH values and residual sugars following fermentation of P72N in a 250mL flask using MM2 at 37 °C for 24 h with different agitation speeds are presented in Supplementary Table 4. A cell number of 9.21±0.07 log CFU/ml with a pH value of 4.13±0.01 and residual sugars of 5.24±0.07 was obtained after a shorter fermentation period of 12 h at an agitation speed of 120 rpm (Fig. 3A and Supplementary Table S4). Consequently, in all further scale-up experiments, an agitation speed of 120 rpm was selected for use.

Fermentation in a 500mL flask

When P72N was grown in a 500 mL flask, cell numbers of 9.70±0.02 log CFU/ml were observed at a fermentation time of 12 h in MM2, whilst cell numbers of 9.71±0.07 log CFU/ml were observed in MRS at 15 h fermentation. The pH values and residual sugar content in MM2 were 4.35±0.05 and 4.61±0.81 g/L at 12 h fermentation, respectively, whilst for MRS they were 4.35±0.09 and 11.00±1.00 g/L at 15 h fermentation, respectively (Fig. 3B and Supplementary Table S5). Based on these findings, MM2 and its optimal fermentation conditions (temperature 37 °C, pH 6.5 ± 0.05 and agitation speed 120 rpm) gave a higher cell yield than MRS and were therefore selected as the most suitable combination for scale-up fermentations in a 5 L bioreactor and a 50 L fermenter.

Scale-up fermentation in a 5 L bioreactor

Batch fermentation of P72N was performed using MM2 in a 5 L bioreactor with a working volume of 3.5 L. As shown in Fig. 4A, there was a rapid increase in viable cells detected from 0 to 6 h of incubation. Viable cells persisted at a stationary phase from 9 h to 18 h, and then steadily decreased after 18 h. The highest cell numbers were observed at 12 h of incubation (9.66±0.09 log CFU/ml). The pH value of the medium markedly declined and reached 4.13±0.02 at 12 h of fermentation, after which the values were not significantly different until the end of fermentation (24 h). The concentration of sugar in the medium dropped markedly to 1.51±0.06 g/L at 12 h of fermentation. After 12 h, the values were not significantly different until the end of fermentation (24 h) (Supplementary Table S6).

Fig. 4
figure 4

Fermentation of P72N in a 5 L bioreactor (A) and in a 50 L fermenter (B) using the modified medium MM2 at 37 °C for 24 h with an agitation speed of 120 rpm, showing cell numbers, residual sugars and pH values.

Scale-up fermentation in a 50 L fermenter

Using MM2, batch fermentation of P72N was carried out in a 50 L fermenter with a working volume of 35 L. As shown in Fig. 4B, a prompt increase in viable cells was detected from 0 to 6 h of incubation. Viable cell numbers remained at a stationary phase from 9 h to 18 h, and then steadily decreased. The highest cell numbers (9.79±0.01 log CFU/ml) were observed at 12 h of incubation. The pH value of the medium was reduced to reach 4.14±0.00 at 12 h of fermentation, after which values were not significantly different until the end of the experiment. The levels of sugar in the medium gradually reduced to reach 3.34±0.24 g/L at 12 h of fermentation. After 12 h, none of the values were significantly different up to the end of the fermentation (Table S6).

Overall, the use of MM2 for P72N provided nearly the same yields when scaling up to 5 L and 50 L. A shorter fermentation time of 12 h was the most suitable harvest time, with the maximum number of cells being over 9.60 log CFU/ml (Table 3).

Table 3 Fermentation of P72N in a 5 L bioreactor and a 50 L fermenter using the modified medium MM2, showing the best harvest time, cell yields, residual sugars and pH values.

Cost comparison

The costs per litre of MM2 and commercial MRS media are compared in Table 4, with the cost of the ingredients available in Thailand calculated in US $ in April 2024. The cost of 1 L of commercial MRS medium ranged between $5.94 and $14.34 per litre (depending on the manufacturer), while the cost per litre of MM2 was $1.95 per litre. Hence, the cost of the production medium could be reduced by approximately 67–86% compared to the commercial MRS medium.

Table 4 Cost comparisons between modified medium MM2 and commercial MRS medium.

Stress tolerance of P72N grown in MM

P72N cells grown in MM were significantly more tolerant to heat and bile challenges than the cells grown in MRS, but oxidative tolerance did not differ between cells grown in MM or MRS. The cells in MM were significantly more sensitive toward acid challenge than the cells in MRS, but the cells grown in MM2 were significantly more tolerant to acid stress than the cells grown in MM1, while the heat, oxidative and bile tolerance did not differ (Fig. 5).

Fig. 5
figure 5

Differences in stress tolerance of P72N grown in commercial MRS medium and modified media MM1 and MM2. Error bars represent the standard deviations for triplicate experiments. a−cValues with different superscripts for each stressor are significantly different (p < 0.05).

Metabolomic analysis

Whole metabolite profiles detection by multivariate statistical analysis

The whole metabolite profiles detected by UHPLC-ESI-Q-TOF-MS were analyzed by multivariate statistical analysis, including principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA), to visualize metabolomic changes before and after P72N culture in the modified medium (MM2) and MRS medium. In Fig. 6A, the PCA score plots depict clear separations between modified medium before (MM2_B) and after (MM2_P) P72N culture, while Fig. 6B shows distinct clustering for MRS medium before (MRS_B) and after (MRS_P) P72N culture, indicating significant metabolomic alterations induced by P72N. Further analysis using supervised PLS-DA models (Fig. 6C and D) highlighted specific differences between the growth media pre- and post-culture. The R², Q², and accuracy values exceeded 0.9, confirming the model’s reliability and excellent predictive performance. These results demonstrated the validity of the PLS-DA models, enabling the extraction of reliable variable importance in projection (VIP) values from these models.

Fig. 6
figure 6

Principal component analysis (PCA) score plots of modified media (A) before P72N culture (MM2_B) vs. modified media after P72N culture (MM2_P); (B) MRS media before P72N culture (MRS_B) vs. MRS media after P72N culture (MRS_P); (C) Partial least squares discriminant analysis (PLS-DA) score plots of MM2_B Vs MM2_P; (D), MRS_B vs. MRS_P.

Differential metabolites identification

Differentially expressed metabolites were identified based on results of a Student’s t-test (p < 0.05) and a log2 fold change threshold of < -2 or > 2, with results visualized in volcano plots (Supplementary Figure S2). Between MM2_B and MM2_P, 331 metabolites were significantly altered, with 154 upregulated and 177 downregulated (Figure S2A). For MRS_B vs. MRS_P, 763 metabolites were significantly different, including 377 upregulated and 386 downregulated metabolites (Figure S2B). The most significantly altered metabolites were identified using Variable Importance in Projection (VIP) values from the PLS-DA model, with a cutoff of VIP > 2. From MM2_B vs. MM2_P, 136 metabolites with VIP > 2 were identified, comprising 83 from MM2_B and 53 from MM2_P, while for MRS_B vs. MRS_P, 100 metabolites with VIP > 2 were identified, including 64 from MRS_B and 39 from MRS_P. The list of these significantly altered metabolites is presented in Supplementary Table S7. The modified medium (MM2_B) predominantly contained peptides, followed by nucleic acids, vitamins, lipids, and other compounds. These include valyltryptophan, adenine, dethiobiotin, etc. After 12 h of P72N cultivation, bioactive metabolites, such as indolelactic acid, 1,4-dihydroxy-2-naphthoic acid, and 2’,4’-Dihydroxy-7-methoxy-8-prenylflavan etc. were found in the medium (MM2_P), as shown in Table S7. Similarly, MRS medium (MRS_B) mostly included peptides, lipids, vitamins and other compounds, such as isoleucyl-valine, CL(i-16:0/a-17:0/18:2(9Z,11Z)/i-24:0), dethiobiotin and isoniazid alpha-ketoglutaric acid etc. After 12 h of P72N cultivation, bioactive metabolites, including indolelactic acid, 1,4-dihydroxy-2-naphthoic acid and diacetyl etc. were found in the medium (MRS_P). As expected, the metabolomic analysis revealed a notable shift in metabolite profiles as the growth medium changed. Bioactive compound production differed. In the modified medium, P72N exclusively induced production of iodoantipyrine, 2’,4’-dihydroxy-7-methoxy-8-prenylflavan, quinidine N-oxide, vicriviroc, bortezomib, lomibuvir, and valnemulin. In contrast, it produced diacetyl, methylisopelletierine, and asiaticoside only in the MRS medium. Notably, the modified medium supported the production of a broader diversity of bioactive metabolites, highlighting its potential to enhance probiotic functionality and its suitability as an alternative medium.

Heat maps of identified metabolites

Changes in the levels of metabolites identified in the growth medium between pre- and post-culture were visualized according to heat maps. Heat maps of the top 50 metabolites of MM2_B vs. MM2_P and MRS_B vs. MRS_P based on VIP scores > 2 were drawn to show changes in metabolite concentrations (Fig. 7). In Fig. 7A, it can be clearly seen that the levels of metabolites in modified media (MM2_B) were significantly different from those of MM2_P after culturing P72N for 12 h. The MM2_B group exhibited a high number of dipeptides, including phenylalanylproline, phenylalanylarginine, valyltryptophan etc., which constituted much of its composition, while certain compounds such as L-arginine, ethyl aconitate, Fe (II)-nicotianamine and niacinamide were also present. After 12 h cultivation of P72N, the above-mentioned compounds from the MM_B group were significantly downregulated in the MM2_P group. Instead, other bioactive metabolites such as indolelactic acid, 1,4-Dihydroxy-2-naphthoic acid, isopentenyl pyrophosphate etc. were significantly upregulated in the MM2_P group. Figure 7B shows that the levels of metabolites in MRS_B medium differed significantly from those in MRS_P. The MRS_B group demonstrated a high number of dipeptides, including isoleucyl-valine, threoninyl-valine, leucylproline etc., which comprised most of its composition. Additionally, it contained smaller amounts of L-arginine, Isoniazid alpha-ketoglutaric acid and other compounds. After 12 h of cultivating P72N in MRS medium, the levels of dipeptides from the MRS_B group were significantly downregulated in the MRS_P group, except for phenyl-leucine, which was notably upregulated in the MRS_P group. Apart from phenyl-leucine, other bioactive metabolites including diacetyl, indolelactic acid, 1,4-dihydroxy-2-naphthoic acid etc., were significantly upregulated in the MRS_P group.

Fig. 7
figure 7

Heat maps of the top 50 metabolites for (A) MM_B vs. MM_P and (B) MRS_B vs. MRS_P based on VIP scores > 2. Each column represents a sample, each row represents a metabolite, and the colour indicates the relative expression of that metabolite in that group of samples.

Detection of metabolites beneficial for gut health

Box plots depicting the production of lactic acid, butyric acid, 1,4-dihydroxy-2-naphthoic acid (DHNA), indolelactic acid (ILA), and diacetyl by strain P72N are presented in Fig. 8. In this study, strain P72N was shown to produce lactic acid, butyric acid, DHNA, and ILA in both MRS and the modified media. However, the levels of lactic acid and butyric acid were consistently lower than those of DHNA and ILA in both media. Compared to MRS medium, P72N produced a higher amount of butyric acid but a lower amount of lactic acid in the modified medium. Furthermore, the modified medium promoted a higher production of DHNA and ILA, demonstrating its effectiveness for P72N production. However, diacetyl production by P72N was observed exclusively in MRS medium.

Fig. 8
figure 8

Box plots of lactic acid, butyric acid, 1,4-dihydroxy-2-naphthoic acid, indolelactic acid and diacetyl produced by P72N in the media before and after incubation.

Discussion

The most crucial components of a culture medium for bacterial cell production are carbon and nitrogen sources. By meeting these and other nutritional requirements of the strain, it should be possible to achieve optimal growth10. LAB species and even strains have been shown to have preferences for different sugars and different optimal concentrations. Thus, it is important to choose the right source of carbon and energy since these may affect the growth and metabolic activity of LAB19. The results of the carbon source screening experiment highlighted the significant impact of different carbon sources on the cell production of P72N, with clear preferences. P72N demonstrated a preference for dextrose as its primary carbon source, with the highest viable cell count occurring at 20 g/L dextrose. Hence P72N can effectively utilize dextrose, likely due to its specific enzymatic machinery that favours the metabolism of this sugar over others19.

LAB species are fastidious microorganisms and cannot grow well on simple mineral media with only a carbon source. Therefore, it is necessary to provide them with a required nitrogen source for biomass overproduction11. Numerous studies have indicated that cell biomass production of most LAB species can be increased in the presence of meat and yeast extract, amino acids, protein concentrates, hydrolysates, vitamins, and inorganic compounds such as (NH4)2SO4 and (NH4)2HPO4 20,21. The nitrogen sources used in the current study were sweet whey, beef extract and yeast extract which are all food-grade, cheap and easily accessible in Bangkok. Sweet whey is a major by-product of the dairy industry, containing lactose, protein, lipids and mineral salts, which are appropriate to supply growth factors for lactose-utilizing LAB strains21. Beef extract contains a diverse array of nutrients, such as amino acids, peptides, vitamins, and minerals, all of which are crucial for microbial growth and metabolism22. Yeast extract is known for its high content of B vitamins, free amino acids, and peptides, which are vital for LAB biosynthetic and energy-yielding pathways23,24. The current study revealed yeast extract was the only nitrogen source that demonstrated a statistically significant effect on viable cell production. This indicates that P72N relies heavily on the complex nutrients and growth factors present in yeast extract, which include amino acids, peptides, and various vitamins that support robust cell growth23,24.

The use of RSM identified optimized media formulations for P72N and demonstrated the importance of balancing nutrient supply with cost considerations. The use of cost-effective ingredients like yeast extract, combined with appropriate concentrations of carbohydrates and essential minerals, ensures high cell viability and robust growth. Since both PBD and RSM with CCD are statistical based, the fitness of the model can be checked by examining R2 values. Usually, the R2 value lies between 0 and 1, and at values closer to 1 the model can predict the response better. A model with an R2 > 0.75 is regarded as acceptable25. The model’s accuracy can also be confirmed by a lack of fit that is statistically insignificant (p > 0.05)26. In the current study, R2 values and the lack of fit values of PBD and CCD experiments were all in acceptable agreement, indicating the model’s fitness and accuracy. Based on the RSM experiment, a medium formulation including 20 g/L dextrose, 45 g/L yeast extract, 1 g/L Tween 80, 2 g/L C6H14N2O7, 2 g/L K2HPO4, 5 g/L CH3COONa, 0.1 g/L MgSO4.7H2O, and 0.05 g/L MnSO4.H2O was initially identified as a suitable modified medium (MM1) for P72N.

In industrial fermentation processes, such as those used in probiotic production, sugars serve as the primary energy source for microorganisms. If these sugars are not fully utilized during fermentation, they can remain as residuals in the fermentation medium, leading to several adverse consequences such as wastewater contamination, microbial imbalance, and soil nutrient disruption27. High carbon source ratios can lead to unutilized sugars which affect product quality and contribute to environmental pollution27. Therefore, to reduce the sugar residues found to be associated with fermentation in MM1, a final medium formulation (MM2) containing 10 g/L dextrose, 45 g/L yeast extract, 1 g/L Tween 80, 2 g/L C6H14N2O7, 2 g/L K2HPO4, 5 g/L CH3COONa, 0.1 g/L MgSO4.7H2O, and 0.05 g/L MnSO4.H2O was chosen for further scale-up experiments.

The results from the fermentation experiments in 250 mL flasks using MM2 provided valuable insights into the optimal agitation speed for maximizing the growth of P72N. Acceptable cell counts were obtained at an agitation speed of 120 rpm after a shorter fermentation period of 12 h. The success of this low agitation speed suggests that P72N prefers relatively low oxygen transfer and low shear stress. This could be attributed to the strain’s metabolic characteristics and its ability to thrive in less turbulent environments28,29. Proper agitation not only improves nutrient availability and waste removal but also influences the physiological state of the cells, which is critical for achieving high cell densities30. The fermentation experiments in 500 mL flasks revealed significant improvements in cell growth and fermentation efficiency using the modified medium. MM2 enabled higher cell yields at a fermentation period of 12 h, with better substrate utilization and acid production compared to the commercial MRS medium. Consequently, MM2 was chosen for further scale-up in 5 L bioreactors and 50 L fermenters, paving the way for more efficient industrial-scale production of strain P72N. The results from scale-up fermentation experiments indicated that higher cell yields were observed at a fermentation period of 12 h in both 5 L bioreactor and 50 L fermenters, with acceptable substrate utilization and acid production, hence providing a shorter and efficient fermentation process. The observed pH levels and residual sugars at the end of the fermentation period are indicative of the metabolic activity of the LAB strain, primarily its production of lactic acid. Maintaining an optimal pH is crucial as extreme pH levels can inhibit bacterial growth and metabolic functions. The slightly acidic pH values are typical for LAB fermentation and reflect efficient sugar utilization and lactic acid production, which is a desired trait for many industrial applications31. Overall, the modified medium proved effective for growing strain P72N at different scales, supporting high cell yields and efficient substrate utilization. The optimal fermentation times identified will guide future large-scale production, ensuring consistent and cost-effective manufacturing of this strain.

A probiotic product must withstand various stressors encountered during manufacturing, storage, transportation, and passage through the gastrointestinal tract. Tolerance assays typically assess resistance to common stress factors such as heat, oxidative stress, bile exposure, and low pH levels, reflecting the challenges faced during processing and digestion2,32. The probiotic strain P72N was initially selected in our previous study based on its tolerance to acid and bile stress5. However, alterations in the growth environment and medium composition can influence probiotic viability, stress resistance, and the synthesis of beneficial compounds. P72N cells grown in MM1 and MM2 exhibited better resistance to heat, and bile stress compared to those grown in commercial MRS medium, with similar oxidative tolerance. However, they were notably more sensitive to acidic conditions. The underlying mechanism of acid sensitivity involves lactic acid accumulation, damage to cell membranes, impaired transport systems, and hindering of the production of stress proteins and an effective acid tolerance response. Nutrient limitations further exacerbate enzyme function, increasing acid sensitivity33,34. In part, this potential limitation of using modified media is likely to be overcome by microencapsulation of the strain before its application8.

Metabolites represent the final products of cellular regulatory processes, with their levels regulated in response to environmental changes within biological systems35. In this study, metabolomic profiling demonstrated a significant shift in the metabolite composition of our probiotic strain P72N when cultured in modified media compared to standard MRS medium. The modified media enhanced the production of a broader range of beneficial metabolites, emphasizing its potential to improve the biofunctional properties of probiotics as well as its suitability as an alternative media. This study demonstrated that P72N produced lactic acid, butyric acid, 1,4-dihydroxy-2-naphthoic acid (DHNA), and indolelactic acid (ILA) in both MRS and modified media. However, the concentrations of lactic acid and butyric acid were consistently lower compared to those of DHNA and ILA. Lactic acid is produced by fermentation of lactic acid bacteria through a metabolic process that converts glucose or other sugars into lactic acid and cellular energy. P72N may use dextrose as the primary carbon source and ferments it primarily via the homofermentative Embden-Meyerhof-Parnas (EMP) pathway to produce lactic acid as the main end product. Lactic acid can help keep the gut acidic, creating an environment that is hostile to pathogens. However, too much lactate can harm gut health36. Butyric acid is a type of short chain fatty acid (SCFA) which helps to preserve the intestinal barrier, inhibit inflammation, and modulate intestinal flora, ultimately contributing to improved intestinal health37. It supports gut health by serving as the primary energy source for colonocytes and strengthening the intestinal barrier. It also exhibits anti-inflammatory effects by modulating immune responses and inhibiting inflammatory signalling pathways such as NF-κB. Additionally, butyric acid promotes a healthy gut microbiota and helps suppress pathogenic bacteria, contributing to overall intestinal homeostasis. In this study, strain P72N produced DHNA and ILA in both media, but their abundance was higher in the modified medium, demonstrating its effectiveness and suitability for P72N production. Although low levels of lactic and butyric acids could limit some traditional probiotic functionalities, the elevated production of DHNA and ILA in our modified medium presents a novel metabolic profile with alternative benefits for gut health. DHNA, derived from bacterial fermentation, is a novel type of prebiotic balancing intestinal flora38, and attenuating colonic inflammation through immunomodulation39. It promotes the growth of Bifidobacterium spp., which enhances microbial diversity and reinforces gut barrier integrity. It also modulates immune responses by suppressing pro-inflammatory cytokines and regulating NF-κB signaling, thus attenuating colonic inflammation. ILA is a beneficial gut bacterial metabolite known for reducing inflammation, correcting microbial dysbiosis40, and inhibiting pathogens41. DHNA is an intermediate in the biosynthesis of menaquinone (vitamin K2) through the shikimate pathway. In LAB, this pathway starts from chorismate, which is converted through a series of enzymatic steps involving menF, menD, menH, and other men genes to produce DHNA. The increased DHNA levels in MM2 suggest that the modified medium may favour the flux through the menaquinone pathway, potentially due to an abundance of precursor metabolites like chorismate or reduced competition for pathway intermediates. ILA is derived from the tryptophan metabolism pathway. In lactic acid bacteria, tryptophan is transaminated by aromatic amino acid aminotransferases (ArATs) into indolepyruvate, which is then reduced by indolelactate dehydrogenase to ILA. The elevated ILA levels in MM2 suggest that the medium contains either higher levels of tryptophan or components that upregulate the tryptophan metabolism pathway. Metabolites like DHNA and ILA may also play intracellular roles in quenching reactive oxygen species or modulating membrane integrity, indirectly enhancing stress resilience. In contrast, the acid sensitivity observed in MM2-grown cells could stem from lower buffering capacity of the medium or lack of adequate amino acid-derived pH homeostasis. In this study, diacetyl, a natural flavour compound produced by LAB strains through citrate metabolism, was only synthesized by P72N in MRS medium. The absence of diacetyl production in MM2 likely results from the absence or low concentration of citrate. Citrate acts as a substrate for citrate lyase, leading to acetoin and diacetyl synthesis. A study suggested that citrate supplementation significantly enhances diacetyl production in certain LAB42. Based on this, incorporating citrate into the modified medium is recommended to boost diacetyl production, thereby improving the flavour profile of fermented products.

Conclusions

This study involved a step-by-step optimization process for creating a cost-effective modified medium for industrial-scale production of the Thai native probiotic P. acidilactici 72 N. By substituting the ingredients of commercial MRS media with food-grade substrates, we successfully produced our probiotic strain in industrial-scale with higher cell yields while significantly lowering the cost of the production media approximately 67–86% compared to commercial MRS media. In addition, cells grown in the modified media showed good responses to stresses that may occur in the manufacturing, distribution and use of the probiotic strains. These evaluations provide insights into the probiotic strain’s adaptation and resilience to environmental changes, enhancing our understanding of its physiological responses and metabolic capabilities under optimized growth conditions. The utilization of cost-effective modified media was associated with a significant increase in the production of the beneficial metabolites for gut health such as 1,4-dihydroxy-2-naphthoic acid and indolelactic acid, emphasizing the efficacy of the tailored media in preserving and enhancing the probiotic functionality. This study outlines a cost-effective medium optimization strategy that, unlike most existing approaches, ensures the preservation of key probiotic traits—addressing a critical but often overlooked gap in current probiotic production research. Exploring the scalability of this cost-effective food-grade modified medium to large-scale is a topic for further investigation. Future research could investigate the effectiveness of encapsulated probiotic P72N for in vivo use as a feed additive for use in pig farms and feed processing industries.

Materials and methods

LAB source and Preparation of cell suspensions

Pediococcus acidilactici strain 72 N (P72N) was obtained from the culture collection at the Faculty of Veterinary Science, Chulalongkorn University. It was originally isolated from the feces of Thai indigenous pigs. It have been suggested as potential probiotic candidates for pig feed supplementation3. The cells were maintained in MRS broth with 20% glycerol at -20 °C. They were reactivated by plating on MRS agar and incubating for 48 h at 37 °C. Between 5 and 10 colonies of P72N from freshly cultured plates were sub-cultured in 10mL MRS broth and incubated at 37 °C under static conditions. After incubation for 18 h, the cell culture was centrifuged at 5000 rpm for 10 min, the supernatant was discarded, and the cell pellets were washed twice with 0.85% normal saline. Cells were resuspended in normal saline and the optical density (OD) value was adjusted to 0.8 ± 0.05 at a 600 nm wavelength to obtain a cell density equivalent to 108 colony-forming units per millilitre (CFU/mL). This cell suspension was used as a basic inoculum for further experiments.

Optimization of media

One-Variable-at-A-Time (OVAT) method

The OVAT method was employed to study the effects of carbon sources on bacterial cell production. This method can assess the impact of each variable individually, but it does not account for interactions between variables26. Each of food-grade carbon sources including glucose (C6H12O6), dextrose (C6H12O6.H2O) and sucrose (C12H22O11) were prepared at concentrations of 10 g/L, 20 g/L, and 30 g/L, each supplemented with 10 g/L yeast extract. The minerals and buffering agents including 1 g/L Tween 80, 2 g/L ammonium citrate (C6H14N2O7), 2 g/L di-potassium hydrogen phosphate (K2HPO4), 5 g/L sodium acetate (CH3COONa), 0.1 g/L magnesium sulphate heptahydrate (MgSO4.7H2O), and 0.05 g/L manganese sulphate monohydrate (MnSO4.H2O) were added to each medium formulation as the same concentrations derived from MRS medium. The initial pH values of all media were set to 6.5 ± 0.05 using 1 M sodium hydroxide (NaOH) and 1 M HCl hydrochloric acid (HCl) before sterilization. About 1% (v/v) of inoculum from the cell suspension ( 108 CFU/mL) of P72N was inoculated into a 15 mL screw cap tubes containing 10 mL of sterilized medium. After incubating the tubes at 37 °C for 24 h under static conditions, viable cell counts were determined using the drop plate method. The experiment was performed in triplicate. Carbon sources that significantly enhanced viable cell production were selected for further experimentation.

Plackett-Burman design (PBD)

The PBD was performed to identify significant nitrogen sources influencing viable cell production of P72N. The design included a mixture of three variables (beef extract, yeast extract and sweet whey). Plackett-Burman statistics were used to set the concentration levels of variables, resulting in a design matrix table with experimental runs and coded values (+ 1, -1) representing different concentrations. Media were prepared based on these design matrix table, incorporating the most suitable carbon source for P72N (dextrose) as determined from the previous experiment. The minerals and buffering agents were added to each formulation as the same concentrations derived from MRS medium. Before sterilization, the initial pH value of all media was adjusted to 6.5 ± 0.05. About 1% (v/v) of inoculum was added to 10 mL of each medium formulation and incubated at 37 °C for 24 h under static conditions. Viable cell counts were measured using the drop plate method, with experiments conducted in triplicate. The following first-order polynomial equation was used to perform mathematical modelling: where Y is the response (cell count), β0, βi and Xi are the model intercept, regression coefficients and independent variables, respectively.

$$\rm{Y = {\beta _0} +\sum {\beta _i}{X_i}}$$
(2)

The most suitable nitrogen sources that positively influenced viable cell production with a confidence level exceeding 95% were selected for further experimentation10,26.

Response surface methodology (RSM)

RSM with central composite design (CCD) was employed to optimize the concentrations of medium components and to assess the effects of each variable and their interactions. Based on the results from the PBD experiment, the most influential carbon and nitrogen sources (dextrose and yeast extract) were used as independent variables in this experiment. After determining the concentration levels of each independent variable, CCD provided a design matrix table with experimental runs and coded values representing different concentrations of each variable. Media were prepared according to these design matrix table, incorporating the same concentrations of minerals and buffering agents. The initial pH values of all media were adjusted to 6.5 ± 0.05 before sterilization. Approximately 1% of inoculum was added to 10 mL of each formulation and incubated at 37 °C for 24 h under static conditions. The cell count was determined by the drop-plate method. The experiment was conducted in triplicate. The response viable cell counts were used to develop a regression model, with the quadratic polynomial equation for the variables as follows:

$${\rm{Y}} = {\rm{\beta _0} + \sum {\beta _i}{X_i} }+ \rm{\sum {\beta _{ij}}{X_i}{X_j} + }\rm{\sum {\beta _{ii}}{X_i}^2}$$
(3)

where Y is the response value of the dependent variables; β0, βi, βij, and βii are intercept, linear, interaction, and squared coefficients of the model, respectively; Xi and Xj are the independent variables. The medium formulation that significantly enhanced viable cell production with a confidence level greater than 95% was chosen for further experimentation10,26.

Fermentation in a 250mL flask

To determine the effect of agitation speed (0 rpm, 120 rpm and 200 rpm) on the cell production of P72N, fermentation was performed in a 250mL flask using the optimal modified medium with a working volume of 50mL. The pH value of the modified medium was adjusted to 6.5 ± 0.05 before sterilization. The sterilized modified medium was inoculated with 1% of inoculum, then the flasks were incubated in shaking incubators at 37 °C for 24 h. Viable cell counts, pH values and residual sugars were examined at 6 h intervals. The experiment was performed in triplicate.

Fermentation in a 500mL flask

Fermentation of P72N was undertaken in a 500 mL flask using the optimal modified medium with a 350 mL working volume. The sterilized medium was inoculated with 1% inoculum, then the flasks were incubated at 37 °C for 24 h with a 120 rpm agitation speed. MRS medium was used as a comparison, and fermentation was performed under the same conditions. Cell viability, pH values and residual sugars in the media were measured at 3 h intervals. The experiment was performed in triplicate.

Scale-up fermentation

Batch fermentation of P72N were performed in a 5 L bioreactor (B.E. Marubishi co., Ltd.) and a 50 L fermenter (B.E. Marubishi co., Ltd.). The modified medium was prepared and sterilized at 121 °C for 20 min. The preparation of two consecutive subcultured inoculums for 5 L and 50 L fermentation was as follows: 1% of inoculum of the cell suspensions ( ̴ 108 CFU/mL) was inoculated in the desired amount of sterilized modified medium in a flask and incubated at 37 °C for 18 h with a 120 rpm agitation speed. The flask culture was used as an inoculum for a subsequent fermentation. The total working volume was 3.5 L for the 5 L bioreactor and 35 L for the 50 L fermenter. The sterilized modified medium was inoculated with 1% of the inoculum from the flask culture. The fermentation was performed for 24 h at 37 °C and a pH value of 6.5 ± 0.05 with a 120 rpm agitation speed. Cell viability, pH values and residual sugars were examined at 3 h intervals. The experiment was performed in triplicate.

Stress tolerance test

Stress challenges were conducted to assess the probiotic characteristics of strain P72N cultivated in modified medium (MM). Cultures were subjected to heat, oxidative, bile, and acid stressors following a 24 h incubation in both MRS and MM. CFU counts were conducted before and after each challenge to determine percentage survival. The heat challenge involved placing 2 mL of P72N culture in a 15 mL Falcon tube into a 60 °C water bath for 10 min43. For the oxidative challenge, 1.25 mM hydrogen peroxide was added to 2 mL of P72N culture and incubated for 1 h at 37 °C44. Acid challenge was induced by centrifuging 2mL of P72N culture at 6,000 × g for 10 min, then resuspending in the same growth medium adjusted to pH 2.5 using HCl, followed by a 1 h incubation at 37 °C45. Bile challenge was undertaken by inoculating 1% from a P72N cell suspension into 2mL of MRS and MM supplemented with 1% (w/v) bile salts (Oxgall, Sigma-Aldrich, Australia), then incubating at 37 °C for 24 h5. The survival rate (%) in the stress tolerance assays was calculated as: (viable cell count after challenge/ viable cell count before challenge) x 100 46.

Metabolomic analysis

About 1% inoculum (3.5 mL) from the cell suspension (~ 10⁸ CFU/mL) of P72N was cultured in a 500 mL flask containing 350 mL of either MM2 or MRS media. The flasks were incubated at 37 °C for 12 h in a shaking incubator with an agitation speed of 200 rpm. After incubation, cell-free supernatants (CFS) were obtained via centrifugation. About 2 mL of each CFS was mixed with 4 mL of cold acetonitrile (-20 °C) for protein precipitation in an ice bath for 20 min. Following centrifugation (11,000 rpm, 10 min, 10 °C), the supernatants were transferred to Eppendorf tubes and concentrated using a vacuum concentrator (30 °C for 2.5 h). The dried samples were stored at -80 °C until further analysis. Metabolomic analysis was performed using an ultra-high performance liquid chromatography coupled with electrospray ionization quadrupole-time-of-flight-mass spectrometry (UHPLC-ESI-Q-TOF-MS) (Bruker’s Compact), with a C18 column (2.1 × 100 mm, 2 μm) at 55 °C and an autosampler set at 10 °C. Mobile phases consisted of water with 0.1% formic acid (A) and methanol with 0.1% formic acid (B), with a gradient flow rate of 0.4 mL/min. Samples (4 µL) were injected for analysis in positive ionization mode, using 70% methanol as a blank. After every five samples, methanol blanks and a QC sample were analyzed to ensure reproducibility and prevent cross-contamination. Sodium formate was used as an external calibrant. Data acquisition included a mass range of 50–1300 m/z under specified ionization conditions (e.g., capillary voltage: 4500 V; dry temperature: 200 °C; nebulizer: 0.5 bar). Multivariate data analyses, including PCA, PLS-DA, and Student’s t-test, were performed using MetaboAnalyst 5.0 to identify differential metabolites in CFS between pre- and post-culture of P72N.

Statistical analysis

Data were presented as the mean value ± standard deviation (SD). PBD and RSM were performed using Minitab version 20. One-way analysis of variance (ANOVA) was used to investigate the effects of variables on the viable cell production, pH values and residual sugars. Values were considered significant at p < 0.05. Duncan’s multiple range test (DMRT) was used to determine the degree of significant differences between mean values. All analyses were performed using SAS version 9.