For initial screening of GABA-producing capacity, the five selected L. plantarum strains (W1, W3, W4, W5 and W12) were cultured in MRS broth supplemented with 1 % (m/V) MSG at 37 °C for 24 h.

To determine the optimal MSG concentration for GABA production by L. plantarum W12 in soymilk, MSG was added at 0, 0.5, 1.0, 1.5 and 2.0 % (m/V). Fermentation temperature was increased to 45 °C for final optimisation to reduce fermentation time to 15 h while maintaining bacterial viability. GABA concentration was measured by high-performance liquid chromatography (HPLC; Shimadzu Scientific Instruments, Columbia, MD, USA).

To evaluate the effect of different carbon sources on GABA production, soymilk supplemented with 1.5 % (m/V) MSG was further supplemented with 5 % (m/V) of individual carbon sources: lactose, sucrose, glucose, or maltose. A control without additional carbon source was also prepared. Each supplemented soymilk was inoculated with L. plantarum W12 and fermented at 43 °C for 15 h. GABA concentration was determined by HPLC (Shimadzu Scientific Instruments).

Based on the carbon source screening results, the optimal sucrose amount was investigated. Soymilk containing 1.5 % (m/V) MSG was supplemented with sucrose at 0, 5, 10, 15 and 20 % (m/V). Fermentation conditions and GABA analysis were performed as described above.

For final product characterisation, L. plantarum W12 was inoculated into soymilk containing optimised concentration of MSG (1.564 mg/mL) and amount of sucrose (10.93 %, m/V) at an initial cell value of 5·10⁶ CFU/mL. Fermentation was conducted at 43 °C with an initial pH=6.0. Samples were collected at 0, 3, 6, 9, 12, 15 and 18 h for analysis of cell growth, pH, GABA concentration, and organic acid production.

Two different sample preparation methods were used depending on the analyte and sample matrix. For fermented soymilk, the extraction procedure was adapted from Costa et al. (17). Briefly, 4 g were accurately weighed and mixed with 5 mL of HPLC-grade water (Sigma-Aldrich, Merck). This mixture was then further diluted with 25 mL of HPLC-grade acetonitrile (Sigma-Aldrich, Merck). The resulting solution was filtered through a 0.45-µm membrane filter to remove any particulate matter. An aliquot of 500 µL of the filtrate was then used for HPLC analysis.

For MRS broth cultures, LAB strains were cultivated in MRS broth. The supernatants were collected after centrifugation at 10 000×g and 4 °C for 15 min (centrifuge 5424R; Eppendorf AG). Proteins were precipitated by adding 3 % (m/V) sulfosalicylic acid, followed by a second centrifugation step. The resulting protein-free supernatant was then used for GABA derivatisation, as described in section GABA quantification.

Organic acids were analysed using an HPLC system (Shimadzu Scientific Instruments) equipped with an OA-1000 organic acid column (6.5 mm×300 mm; Alltech Associates Inc., Columbia, MD, USA) and a UV detector set at 210 nm. The mobile phase consisted of 0.005 M H₂SO₄ at a flow rate of 0.4 mL/min, and the column temperature was maintained at 35 °C. Reference organic acids (Sigma-Aldrich, Merck) were used for calibration and quantification.

GABA was analysed following a method adapted from Thuy et al. (18). Briefly, GABA in the prepared samples was derivatised with 4 mM 4-dimethylaminoazobenzen-4’-sulfonyl chloride (dabsyl chloride) for soymilk and MRS broth samples. Dabsyl-GABA was quantified at λ=465 nm using a Shimadzu LC-20A HPLC system equipped with an SPD-20A UV-Vis detector. Separation occurred on a Supelco C18 column (250 mm×4.6 mm i.d., 5 μm particle size) maintained at 55 °C. The mobile phase consisted of an isocratic mixture of 25 mM ammonium acetate (0.1 % acetic acid) and acetonitrile (φ(acetonitrile)=74 %) at a flow rate of 1 mL/min. GABA concentrations were determined using a calibration curve in the range 0–10 mM.

LAB cells were harvested from overnight cultures by centrifugation (10 000×g, 4 °C, 15 min, centrifuge 5424R; Eppendorf AG) and resuspended in a sterile solution of 0.1 % Bacto peptone (Difco Laboratories) and 0.9 % (m/V) NaCl. A volume of 1 mL of cell suspension was then inoculated into MRS broth and adjusted to pH=2.5. The initial absorbance (A_T0) was measured at 600 nm. Following incubation at 37 °C for 4 h, the absorbance was measured again (A_T4). Survival percentage (H/%) was calculated using the following equation:

LAB cells from overnight cultures in MRS broth were harvested by centrifugation (10 000×g, 4 °C, 15 min, centrifuge 5424R; Eppendorf AG) and resuspended. A volume of 1 mL of cell suspension was inoculated into a sterile saline solution (0.85 % NaCl, m/V) containing 1 mg/mL pancreatin (Sigma-Aldrich, Merck, Seoul, South Korea) and 0.3 % (m/V) bile salts (MilliporeSigma, Merck KGaA, Darmstadt, Germany) The solution was incubated at 37 °C. Absorbance was measured at the beginning (A_T0) and after 6 h of inoculation (A_T6). Tolerance was calculated as survival percentage (H/%), using the following equation:

After 24 h of incubation, LAB cultures were centrifuged (400×g, 4 °C, 6 min, centrifuge 5424R; Eppendorf AG). Simulated gastric fluid was prepared by adding pepsin (Sigma-Aldrich, Merck, Burlington, MA, USA) to a 0.85 % NaCl solution to achieve a final concentration of 3 mg/mL and adjusting the pH to 2.5. The cell pellet was resuspended in the simulated gastric fluid, vortexed, and the initial absorbance was measured at 600 nm (A_T0). After incubation at 37 °C for 4 h, the absorbance was measured again (A_T4). Survival percentage (H/%) was calculated using Eq. 1.

Antibiotic susceptibility was determined using the disc diffusion method of Oh et al. (20) and interpreted according to the guidelines established by the European Food Safety Authority (EFSA) for Lactiplantibacillus species (21). The following antibiotic discs were placed on agar plates containing bacteriological grade agar powder (GRM026; HiMedia, Mumbai, India) inoculated with the lactic acid bacteria (LAB) strains (in µg): vancomycin 30, clindamycin 2, tetracycline 30, ampicillin 10, streptomycin 10, penicillin 10, chloramphenicol 30, erythromycin 15, kanamycin 30 and gentamicin 10. After 24 h of incubation at 37 °C, the diameters of the inhibition zones were measured.

Auto-aggregation: Following the method of Li et al. (19), overnight LAB cultures were centrifuged (10 000×g, 4 °C, 10 min, centrifuge 5424R; Eppendorf AG), and the pellets were resuspended in 1 mL of phosphate-buffered saline (PBS). Initial absorbance (A₀) was measured at 600 nm. After incubation at 37 °C for 20 h, the absorbance of the upper suspension (A₂₀) was measured. Autoaggregation (%) was calculated using the following equation:

Co-aggregation: LAB cells from 24-hour cultures in MRS broth were mixed with equal volumes (1.5 mL) of pathogenic strains (Staphylococcus aureus, Escherichia coli and Salmonella Typhimurium). The mixtures were incubated at 37 °C for 5 h. The absorbance of the supernatants (A_mix) was measured at 600 nm. The absorbance of the individual LAB (A_probiotic) and pathogen (A_pathogen) suspensions was also measured. Co-aggregation (%) was calculated using the following equation:

LAB cultures grown in MRS broth at 37 °C for 24 h were centrifuged (400×g, 4 °C, 10 min, centrifuge 5424R; Eppendorf AG). The pellets were resuspended in 0.1 M KNO₃. The initial absorbance (A₀) was measured at 600 nm. Cell suspensions of 2 mL were mixed with 1 mL of organic solvent (chloroform, ethyl acetate or xylene). After incubation at room temperature for 10 min and vortexing for 2 min, the mixture was left undisturbed for 20 min to allow complete phase separation, and the absorbance of the aqueous phase (A₁) was then measured. Cell surface hydrophobicity (%) was calculated using the equation below according to Oh et al. (20):

The antimicrobial activity was determined using the agar spot method of Hernandez et al. (22). Pathogen indicator strains included S. aureus, S. Typhimurium and E. coli. A suspension of LAB cells was prepared by mixing the overnight LAB biomass in 0.85 % NaCl. This suspension (5 µL) was dropped onto MRS agar surfaces, allowed to dry, and grown at 37 °C for 18 h. Cells of pathogenic bacteria were suspended in 1 % agar at approx. 37 °C. These agar suspensions were immediately poured onto the spot-inoculated MRS agar plates, left to solidify, and then incubated at 37 °C for 24–48 h. The diameters of the clear zones of inhibition of pathogenic bacteria around the spots were determined using a Vernier calliper. Positive controls consisted of MRS agar plates inoculated with known antimicrobe-producing L. plantarum reference strain (ATCC 14917), while negative controls consisted of uninoculated MRS agar spots overlaid with pathogen-containing agar. Inhibition zones were measured only when they exceeded 2 mm outside the LAB colony margin to ensure specific antimicrobial activity.

Survival under simulated gastrointestinal conditions is crucial for probiotic efficacy. All tested strains showed considerable tolerance to low pH, bile salts, pancreatin and pepsin (Table 3). L. plantarum W3 and W12 exhibited the highest survival rates in low pH and bile salt/pancreatin conditions (over 95 and 96 % survival, respectively). While all strains showed good pepsin tolerance (above 84 %), W1 and W12 had the highest survival rates, exceeding 97 %. These results suggest that these strains possess robust resistance to the harsh conditions of the gastrointestinal tract. Acid tolerance is a critical probiotic trait, enabling survival through the stomach. At low pH (<4.5), proton influx requires substantial ATP for homeostasis, potentially leading to metabolic disruption and cell death (28). Gastric pH can drop to 1–2 during fasting (29). Thus, a cutoff of pH=3.0 for 3 h was used to simulate gastric conditions. All assessed L. plantarum strains survived well at pH<3.0. Bile tolerance is also essential for probiotic efficacy. Bile salts exert antimicrobial effects by disrupting bacterial membranes (30).

Autoaggregation, the ability of bacterial cells to clump together, and coaggregation, the ability to adhere to pathogenic bacteria, are important for colonisation and competitive exclusion of pathogens in the gut. Table 3 shows that L. plantarum W12 had the highest autoaggregation ability (96.67 %), significantly higher than the other strains. While W12 also showed strong coaggregation with S. aureus (20.82 %) and S. Typhimurium (14.39 %), W3 demonstrated the highest coaggregation with S. Typhimurium (17.74 %). These varying co-aggregation profiles suggest strain-specific interactions with different pathogens.

Probiotic activity is also linked to the ability to aggregate and adhere to host tissues. Autoaggregation and coaggregation help form a barrier against pathogens (31). These traits were evident in the L. plantarum strains tested, supporting their probiotic potential (32).

Cell surface hydrophobicity affects the adherence of probiotic bacteria to intestinal epithelial cells, contributing to colonisation. Table 3 (hydrophobicity) shows that L. plantarum W12 exhibited the highest hydrophobicity towards chloroform (85.45 %) and ethyl acetate (12.71 %), while W1 showed the highest hydrophobicity towards xylene (9.79 %). The variations in hydrophobicity across different solvents indicate differences in cell surface composition among the strains. Surface hydrophobicity correlates with aggregation and adhesion capabilities (33). Hydrophobicity was particularly high in strains W1 and W12, as assessed in chloroform. Bile exposure reduced hydrophobicity, likely affecting adhesion, which is consistent with previous findings (34).

The ability to inhibit pathogenic bacteria is a key probiotic trait. L. plantarum W3 and W12 (Table 3, antimicrobial activity) showed the strongest antagonistic activity against all tested pathogens (E. coli, Salmonella and S. aureus), as indicated by the largest zones of inhibition. W12 notably exhibited the largest inhibition zones against S. aureus (28 mm) and Salmonella (14.33 mm), while W3 showed the strongest effect against E. coli (8.33 mm). These results highlight the potential of these strains to control the growth of common foodborne pathogens.

Based on the comprehensive evaluation of these probiotic properties, L. plantarum W12 emerged as the most promising candidate. Its superior performance in GABA production, combined with high gastrointestinal tolerance, autoaggregation, coaggregation with specific pathogens, and strong antagonistic activity, justifies its selection for further investigation in soymilk fermentation. This strain has significant potential for developing functional foods with enhanced probiotic benefits. The antibacterial activity of L. plantarum is well-documented (35), primarily through organic acid production and pH reduction (36). W12 showed antagonistic effects against E. coli, S. aureus and S. Typhimurium, likely due to its high GABA output and acidification. All tested strains produced significantly larger inhibition zones than negative controls (no inhibition zone), and positive controls using L. plantarum ATCC 14917 showed comparable inhibition patterns, validating the assay methodology.

The results presented here show that L. plantarum W12 successfully integrates two critical functional properties: exceptional GABA biosynthesis capacity and robust probiotic characteristics. The high survival rates of the strain under simulated gastrointestinal conditions (>96 % for pH, bile/pancreatin, and pepsin challenges, Table 3), strong auto-aggregation (96.67 %), and broad-spectrum antimicrobial activity against foodborne pathogens (inhibition zones of 28.00 mm against S. aureus, 14.33 mm against Salmonella and 3.00 mm against E. coli) provide confidence in its probiotic efficacy. The optimised GABA yield of 34.5 mM represents a 3.8-fold improvement over the initial screening value (9.1 mM) and positions this fermented soymilk among the highest GABA-containing plant-based fermented products reported to date.

The viable cell count exceeding 7.9 log₁₀ CFU/mL at the end of fermentation ensures adequate probiotic dosage (typically ≥10⁷ CFU/mL for claimed health benefits) (37), while the high GABA concentration provides therapeutic potential for neurological and cardiovascular benefits. This dual functionality, combined with favourable sensory properties (mild acidity and clean flavour) and extended shelf life conferred by organic acid preservation, positions this product as a promising candidate for commercialisation as a functional, health-promoting beverage. Future research should focus on in vivo validation of GABA bioavailability and probiotic colonisation efficiency through animal models and clinical trials, as well as shelf-life stability studies and consumer acceptance testing to facilitate commercial development.

The initial cell count (5·10⁶ CFU/mL), temperature (45 °C), initial pH (7) and fermentation time (15 h) were kept constant to investigate the specific effects of MSG concentration and carbon source supplementation on GABA biosynthesis.

Additional MSG concentration in soymilk significantly affected GABA production by L. plantarum W12 (Fig. 1b). GABA production increased with MSG supplementation up to 1.5 % (m/V), reaching a maximum of (21.4±0.6) mM. However, increasing the MSG amount to 2 % did not further increase GABA production and showed a slight decrease, suggesting a potential inhibitory effect at higher concentrations or substrate saturation. This finding agrees with other studies reporting an optimal MSG concentration for GABA production in LAB fermentations (38). The decrease in GABA production at 2 % MSG could be due to substrate inhibition of glutamate decarboxylase or feedback inhibition mechanisms within the metabolic pathway (39). Therefore, 1.5 % MSG supplementation is optimal for maximising GABA production by L. plantarum W12 under these fermentation conditions. While high glutamate concentrations can enhance GAD activity (40), excessive amounts may disrupt metabolism via osmotic stress. Our data suggest that GABA production peaked at 15 h, followed by a decline, possibly due to substrate depletion, GABA degradation, or GAD feedback inhibition (41).

The results (Fig. 1c) indicate that the type of carbon source (lactose, sucrose, glucose or maltose) significantly affected GABA production. Supplementation with sucrose resulted in the highest GABA production ((29.3±0.8) mM), significantly exceeding all other treatments (p≤0.05). Glucose supplementation also led to a substantial increase in GABA ((26.4±0.5) mM), while lactose and maltose supplementation resulted in GABA concentrations comparable to the control (no added carbon source) ((24.3±1.2) and (22.8±0.7) mM, respectively).

The differences in GABA production with different carbon sources can be attributed to variations in their metabolic pathways and how efficiently they support the growth and metabolic activity of L. plantarum W12. Sucrose, a disaccharide composed of glucose and fructose, may provide a more sustained release of readily metabolisable sugars, leading to enhanced GABA production (42). The lower GABA concentrations observed with lactose and maltose could be due to differences in their uptake and utilisation by L. plantarum W12 or potential catabolite repression effects. Selecting an appropriate carbon source is therefore crucial for optimising GABA production in soymilk fermentation by L. plantarum W12.

Next, the optimal sucrose amount for GABA production was investigated. Varying amounts of sucrose (0, 5, 10, 15 and 20 %, m/V) were added to the MRS medium containing 1.5 % MSG. The results (Fig. 1d) of L. plantarum W12 demonstrate a clear influence of sucrose content on GABA production. GABA production increased with up to 10 % sucrose, reaching a maximum of (33.6±1.4) mM. However, further increases in sucrose content (15 and 20 %) resulted in decreased GABA production ((30.8±1.2) and (27.6±0.7) mM, respectively). This suggests that while sucrose enhances GABA production up to a certain point, excessive amounts can have a detrimental effect.

The decrease at higher amounts of sucrose could be attributed to factors including osmotic stress, substrate inhibition, and a shift in metabolic flux. High sucrose amounts can create osmotic stress for the bacteria, inhibiting their growth and metabolic activity (43). Excessive sucrose could potentially inhibit enzymes involved in GABA metabolism or related pathways. High sugar content might redirect metabolic flux away from GABA production towards other metabolic pathways, such as exopolysaccharide production or other stress responses (44). The results indicate that 10 % sucrose supplementation is optimal for maximising GABA production by L. plantarum W12 under these fermentation conditions. This finding is valuable for developing an efficient and cost-effective fermentation strategy for GABA-enriched soymilk production.

The MSG and sucrose amounts for enhanced GABA production by L. plantarum W12 in soymilk were optimised using a central composite design (CCD) of response surface methodology (RSM). The CCD comprised 13 experimental runs, including factorial, axial, and centre points (Table 2). The independent variables were MSG concentration (X₁) and sucrose amount (X₂), with GABA yield (Y, mM) as the response variable. GABA yield varied significantly depending on the concentrations of MSG and amount of sucrose.

A p<0.05 indicates that model terms are significant, while values greater than 0.10 indicate that the model terms are not significant. In the analysis of variance (45), the response surface model optimising GABA yield by L. plantarum W12 in soymilk was developed using a CCD of RSM with MSG concentration (X₁) and sucrose amount (X₂) as independent variables (Table 2). The quadratic model for GABA yield (Y, mM) was highly significant (F=39.65, p<0.0001, R²=96.59 %), with significant terms for MSG (p=0.0338), sucrose (p=0.0284), their interaction (p=0.0319), and quadratic terms for both MSG (p<0.0001) and sucrose (p<0.0001) (Table 4 and Table 5).

The coded values for MSG (X₁) and sucrose (X₂) were converted to actual values using Eqs. 6 and 7.

The following second-order polynomial regression model was fitted to the data:

where Y is the GABA yield (mM), and X₁ and X₂ are the coded values for MSG concentration and sucrose amount, respectively.

Response surface methodology (RSM) was used to visualise and optimise the combined effects of MSG (mg/mL) and sucrose (%, m/V) amounts on GABA yield (mM). The 3D and contour plots (Fig. 2a and Fig. 2b) showed a non-linear relationship between the independent variables and GABA yield, with a clear optimal region indicated by the curved surface and concentric rings, respectively. Model optimisation predicted a maximum GABA yield (Y_max) of 34.2 mM at 1.6 mg/mL MSG and 10.9 % sucrose. These optimised conditions were experimentally validated. The optimisation results were: Y_max=34.2 mM, X₁(MSG)=1.6 mg/mL, and X₂(sucrose)=10.9 %.

A time course study was conducted to monitor changes in pH, cell growth, organic acid production, and GABA concentration during soymilk fermentation by L. plantarum W12 under optimised conditions (initial cell density 5·10⁶ CFU/mL, 45 °C, initial pH=6.0). Key factors influencing microbial GABA biosynthesis include cell density, pH, temperature, fermentation time and glutamate availability (18).

The primary objective of this study was to optimise GABA production by L. plantarum W12 in soymilk. GABA production was monitored throughout the 18-hour fermentation (Fig. 3a). The results show that GABA production is time-dependent. Initially, GABA concentrations were negligible, but they increased steadily, reaching a maximum of (34.5±1.0) mM at 15 h. A slight decrease in GABA concentration was observed at 18 h ((33.5±0.7) mM), although this difference was not statistically significant (p>0.05). The increase in GABA production correlates with the growth of L. plantarum W12 and the decrease in pH. This suggests that active bacterial metabolism and the acidic environment created by organic acid production are favourable for GABA biosynthesis (46).

L. plantarum W12 reached a GABA concentration of 9.1 mM under optimal conditions and demonstrated strong acid and bile resistance, high hydrophobicity, and antimicrobial activity, supporting its suitability for functional food applications. The observed pH decrease (from ~6.0 to 4.6 over 18 h) reflects active metabolism and organic acid production, notably lactic acid, which contributes to preservation and sensory properties (47). At 15–18 h, W12 entered the stationary phase with maximum GABA accumulation. GAD activity, optimal at low pH, may be suppressed as GABA production increases at medium pH. Mutant GAD enzymes with broader pH activity have been proposed to overcome this limitation (48). Recently, Kubota et al. (7) showed that GABA attenuates ischaemia-reperfusion-induced alterations in intestinal immunity through increased IgA secretion, alpha-defensin-5 expression, and small intestinal superoxide dismutase activity. These findings indicate that L. plantarum W12 in our study is a potential probiotic for producing functional foods to enhance intestinal health.

Cell growth is a critical factor in fermentation as it relates to the production of metabolites, including GABA. The growth of L. plantarum W12 in soymilk was monitored over 18 h of fermentation (Fig. 3b). The data show a typical growth curve for L. plantarum W12 under these conditions. The initial cell density of 6.7 log CFU/mL increased rapidly during the first 12 h, reaching 7.8 log CFU/mL. After 12 h, the growth rate slowed, and the cell density reached 7.9 log CFU/mL by the end of the 18-hour fermentation. This indicates that L. plantarum W12 efficiently utilises the soymilk medium supplemented with MSG and sucrose for growth. The observed growth pattern suggests that the bacteria entered the stationary phase after 12 h, due to nutrient depletion or the accumulation of inhibitory metabolites (49). It is important to note that the production of GABA is often associated with bacterial growth and metabolic activity (50). Therefore, understanding the growth dynamics of L. plantarum W12 is crucial for optimising GABA production. The correlation between cell growth and GABA concentration will be further investigated in subsequent analyses. This will help determine the optimal fermentation time for maximising GABA yield while considering the cost and efficiency of the process.

Monitoring pH changes during fermentation is crucial as it reflects the metabolic activity of the bacteria and can influence enzyme activity, including glutamate decarboxylase, which is responsible for GABA production (51). The pH of soymilk decreased significantly throughout the fermentation period (Fig. 3c).

Regarding organic acid profiles, mass fractions of lactic, acetic, pyruvic, propionic and formic acid were monitored over the 18-hour fermentation (Fig. 3d–Fig. 3g). Lactic acid (Fig. 3d) was the dominant organic acid produced, consistent with the metabolic activity of L. plantarum W12, a lactic acid bacterium. Its concentration increased steadily throughout fermentation, reaching a maximum of (182.4±7.9) mg/g at 18 h. This progressive increase in lactic acid is directly related to the observed decrease in pH. Acetic acid (Fig. 3e) was produced at much lower mass fractions than lactic acid, reaching approx. 1.10 mg/g by the end of fermentation. Pyruvic acid mass fraction also increased gradually during fermentation, reaching 0.12 mg/g at 18 h. Pyruvic acid (Fig. 3f) is an important intermediate in various metabolic pathways, including lactic acid fermentation (52). Propionic acid (Fig. 3g) mass fractions remained stable throughout fermentation, around 0.25 mg/g. Formic acid (Fig. 3h) production showed a different pattern, with an initial increase followed by a decrease. The highest mass fraction (0.02 mg/g) was observed at 9 h.

Our results showing lactic acid as the predominant byproduct ((182.4±7.9) mg/g) with minor amounts of acetic acid (1.10 mg/g) align with the typical homofermentative metabolism of L. plantarum. This metabolic pattern is crucial for both product quality and GABA production efficiency. The high lactic acid to acetic acid ratio (165:1) confirms the homofermentative nature of L. plantarum W12, in sharp contrast to heterofermentative LAB such as L. brevis, which produce higher ratio of acetic acid alongside lactic acid and generate additional metabolic byproducts, including ethanol and CO₂ (53).

The dominance of lactic acid in our fermented soymilk offers several advantages: (i) desirable sensory properties, including mild acidity (pH=4.6) and a clean flavour profile without the sharp, vinegary notes associated with high acetic acid content, (ii) effective preservation through controlled pH reduction, inhibiting the growth of spoilage organisms and foodborne pathogens, (iii) favourable metabolic environment for GAD enzyme activity, as the enzyme shows optimal activity at pH=4.5–5.5, which is maintained during the mid-to-late fermentation stages, and (iv) minimal production of off-flavour compounds, as evidenced by the low mass fractions of other organic acids (pyruvic 0.12, propionic 0.25 and formic acid 0.02 mg/g). These low mass fractions of minor organic acids indicate efficient metabolic flux towards lactic acid production, minimising potentially undesirable flavour compounds and ensuring a clean, acceptable taste profile for the final product (54).

The temporal dynamics of organic acid production during fermentation (Fig. 3d–Fig. 3h) showed coordinated metabolic processes: lactic acid accumulation parallelled GABA production and the decrease in pH, suggesting coupling between energy metabolism and glutamate decarboxylation. This relationship is consistent with the GAD-mediated acid resistance mechanism, in which GABA synthesis consumes intracellular protons, helping maintain cellular pH homeostasis during lactic acid fermentation (55). The stable low mass fraction of propionic acid and the transient peak of formic acid at 9 h (followed by a decline) indicate active secondary metabolism without accumulation of undesirable metabolites.