Enhancement of Natural GABA Production in Yogurt by Simple Carbohydrates and Metabolomics Pro ling During Fermentation of Novel, Self-Cloned Lactobacillus Plantarum Taj-Apis362

Farah Hussin Section of Food Engineering Technology, Malaysian Institute of Chemical and Bio-Engineering Technology, Universiti Kuala Lumpur Shyan Chay Department of Food Science, Faculty of Food Science and Technology, University Putra Malaysia Mohd Syahmi Abdul Ghani Department of Food Science, Faculty of Food Science and Technology, University Putra Malaysia Anis Meor Hussin Department of Food Science, Faculty of Food Science and Technology, University Putra Malaysia Wan Wan Ibadullah Department of Food Science, Faculty of Food Science and Technology, University Putra Malaysia Belal Muhialdin Department of Food Science, Faculty of Food Science and Technology, University Putra Malaysia Nazamid Saari (  nazamid@upm.edu.my ) Department of Food Science, Faculty of Food Science and Technology, University Putra Malaysia


Introduction
Yogurt is the fermented form of milk with a thick consistency and has been consumed since ancient time. In modern days, yogurt is appreciated for its high nutritional value and positive health bene ts, owing to the probiotic effects from starter culture, i.e. lactic acid bacteria Streptococcus. thermophilus and Lactobacillus delbrueckii subsp. bulgaricus such as lactose digestion improvement 1 , diarrhea prevention 2 and gut immune system stimulation 3 . Globally, the yogurt market has reached a value of approximately 85.54 billion USD in 2019, and is forecasted to increase to 106.6 billion USD by 2024 4 .
Gamma aminobutyric acid (GABA) is a non-protein amino acid which is widely distributed in animals, plants and microorganisms. In animals, it acts as the primary inhibitory neurotransmitter in central nervous system while in plants and microorganisms, it plays a key metabolic role in Krebs cycle 5 .
Physiologically, GABA has signi cant functions such as stress reduction 6 , inhibition of cancer cell proliferation 7 , blood pressure reduction 8 and prevention of diabetes 9 . The biosynthesis of GABA occurs mainly through the fermentative reaction from microorganisms such as yeast, fungi and bacteria. Most lactic acid bacteria (LAB), namely L. brevis, L. paracasei, L. plantarum, and Lactococcus lactis, have been reported to produce GABA through α-decarboxylation of glutamate substrate via enzymatic reaction of glutamate decarboxylase (GAD), a pyridoxal 5'-phosphate (PLP) dependent enzyme 10 . In this study, selfcloned and expressed L. plantarum Taj-Apis362 recombinant cells; UPMC90 (intracellular) and UPMC91 (extracellular) previously engineered by Tajabadi et al. 11 , were used to improve the GABA production in yogurt and can be considered safe since according to 90/219/EEC 12 , self-cloning is de ned as reintroduction of DNA from a host that has been modi ed, or is closely related to the same species strain, and was excluded from the EU Directive on the contained use of genetically modi ed micro-organisms.
Modi ed organism by self-cloning technique are now not viewed as genetic modi ed organism (GMO) and are regarded safe and suitable for food applications 13 .

Page 3/17
The need for high concentration of glutamate (32-507 mM) as well as the presence of PLP (18-200 µM) have been identi ed as the major obstacles to produce GABA in a food system 14 , 15 as glutamate will produce salty/savoury taste at high concentration that is unfavourable for a yogurt product while PLP cofactor is a costly ingredient. However, due to the various health bene ts of GABA, yogurt rich in GABA represents a value-added functional dairy product that can be conveniently consumed on a regular basis.
To the best of our knowledge, there are only a few studies reported on the effect of sugar on enhancing GABA production in culture medium 16 , 17 and no work has been reported on the effect of prebiotics in culture medium, let alone the effect of these simple carbohydrates in an actual food system. Therefore, the aims of this research were to study : (1) the effect of different types of carbohydrate (simple sugars and commercial prebiotics) on enhancing GABA production in yogurt, cultured by two novel, self-cloned LAB strains (L. plantarum Taj-Apis362, assigned as UPMC90 and UPMC91 by Institute of Bioscience, Universiti Putra Malaysia, Malaysia) in order to mitigate the over-use of glutamic acid and omit the need of PLP cofactor, (2) the stability of GABA-rich yogurt during gastrointestinal digestion and 28-days of refrigerated storage, and (3) the metabolomics pro ling of the fermentation-derived biomolecules in yogurt using 1 H-NMR technique.

Results And Discussion
Effect of Simple Carbohydrates on GABA Production in Yogurt GABA production by microorganism is critically affected by factors including microbial genetic characteristics, culture condition (temperature, pH, time) and medium condition (presence of glutamate and PLP). Most studies reported the requirement of glutamate at high concentrations (32-507 mM) and the presence of PLP (18-200 µM) to achieve optimum GABA production in different food systems 14 , 15 . In an attempt to mitigate glutamate and PLP usage during fermentation, current study is thus developed to improve the medium condition by incorporating different simple carbohydrates, in the form of simple sugars and prebiotics, to maximize GABA production. Our work identi es the lowest amount of glutamate (11.5 mM) as the effective concentration to foster optimum GABA production, as compared to previous works which reported glutamate usage of 32 mM in fermented palm date residue 15 and 80 mM in fermented milk 18 . This proves the e cient conversion of glutamate to GABA in yogurt fermented by UPMC90 and UPMC91 LAB strains under pre-de ned optimum conditions 19 .
The effect of different simple sugars and prebiotics on GABA content and conversion rate is depicted in Fig. 1a. Of all six simple carbohydrates, simple sugars induce higher GABA production (42.83-58.56 mg/100g) compared to prebiotics (34.19-40.51 mg/100g). In particular, glucose signi cantly (p<0.05) induced the highest GABA production in yogurt (58.56 mg/100g, conversion = 34.60%), favourably surpassing control sample added with PLP (48.01 mg/100g, conversion = 28.38%), a cofactor well known to promote GABA biosynthesis. In terms of viable cell count, glucose again signi cantly showed the highest probiotic count (9.31 log CFU/g), followed by sucrose (9.06 log CFU/g) and fructose (8.98 log CFU/g), as depicted in Fig. 1b. The e cient utilization of glucose by LAB strains (both UPMC90 and UPMC91) and starter culture to produce the highest GABA and probiotic count is expected, as glucose is readily phosphorylated to glucose-6-phosphate in the glycolytic cycle of Embden-Meyerhof pathway and phosphoketolase pathway to achieve bacterial cell growth. In contrast, the other two simple sugars, namely sucrose and fructose, have to go through additional conversion steps in phosphoenolpyruvatedependent phosphotransferase system before being converted into pyruvate 20 , which then either splits into the GABA-shunt pathway to form GABA or continues to be decarboxylated through glycolytic pathway to generate ATP, NADH and NADPH for cell growth 21 . The straightforward metabolism pathway of glucose explains the rapid bacterial growth when the sugar is present, allowing the growth cycle to reach an exponential phase in a shorter time compared to other simple carbohydrates. The accumulation of active bacterial cell then contributes towards higher secretion of GAD, thus higher enzymatic activity to convert glutamate into GABA. This observation establishes a positive correlation between viable cell count and GABA production during fermentation, i.e. higher viable count is associated with higher GABA formation.
An interesting observation was made on the yogurt containing prebiotics (inulin, FOS, GOS), of which they recorded GABA contents and viable cell counts that were substantially lower than simple sugars. This indicates the poor utilization of prebiotics during bacterial metabolic activity as a result of high degree of polymerization (DP) ranging from 2 to 65 in the prebiotics 22 . Similarly, Hernandez-Hernandez et al. 23 reported lower growth of L. casei ATCC11578 and L. delbrueckii subsp. lactis ATCC4797 strains when GOS is present instead of glucose and lactulose (a synthetic disaccharide). Rayes 24 also described that carbohydrates with high DP were poor substrates for bi dobacteria. This is because prebiotics with higher DP requires cleavage into monosaccharides with lower DP by extracellular bacterial enzymes prior to transportation into the cells for growth 25 .
Among three tested prebiotics, inulin produced signi cantly higher GABA than FOS and GOS, but lower probiotic growth compared to FOS ( Fig. 1a and 1b), suggesting that LAB strains of UPMC90 and UPMC91, with high GAD enzyme activity, prefer to metabolize inulin over FOS. In accordance to our results, Choudhary et al. 26 also showed that inulin was fermented at a higher rate than FOS by L.
paracasei CD4 in soymilk. According to Sarbini & Rastall 27 , there are speci c transport systems in LAB for trisaccharides and tetrasaccharides, indicating the occurrence of different metabolic capacity based on type of substrate used. In addition, each strain of LAB have their preferred choice of prebiotics to prioritize as substrate during fermentation, depending on their respective genetic characteristics 28 . These ndings explain the preferred utilization of inulin over FOS by the UPMC90 and UPMC91 strains used in current study. On the other hand, the high viable cell count in FOS compared to inulin is mostly contributed by the abundance of starter culture, i.e. S. thermophilus and L. delbrueckii ssp. bulgaricus, present in the sample.
For PLP sample (50 µM, positive control), GABA production is enhanced (48.01 mg/100g, conversion = 28.38%) compared to yogurt without addition of any simple carbohydrates (negative control, 29.95 mg/100g, conversion = 17.71%). This result demonstrated a higher GABA conversion compared to that reported by Yi & Chui 29 , who used 8.09 mM of PLP and 134 mM of monosodium glutamate in adzuki bean milk fermented by L. rhamnosus GG. PLP sample also showed the lowest viable cell count (log 6.76 CFU/g) similar to negative control, indicating no signi cant effect on promoting bacterial cell growth compared to simple carbohydrates. This is supported by previous study by Li et al. 30 who reported that PLP did not affect the growth of L. brevis NCL912. The presence of PLP in yogurt is concluded to enhance GABA production but has no effect in promoting the growth of L. plantarum Taj-Apis362 strain UPMC90 and UPMC91. Among all tested simple carbohydrates, glucose displayed the highest substrate e ciency for bacterial metabolism, rapidly promoting cell growth which enhanced the conversion of glutamate into GABA during yogurt fermentation, thus was selected for further studies as detailed below.

Simulated Digestion Study on GABA-rich Yogurt
In vitro gastrointestinal digestion is regarded as an effective and valid strategy to simulate digestive activities in the human gastrointestinal tract. Various mechanical, chemical and enzymatic actions occur within the human digestive tract to degrade food matrices, releasing nutrients that are readily absorbed by the body. In this study, simulation was performed at pH 1.2 for the rst 2 h to mimic stomach digestion, followed by increasing the pH to 6.8 for the next 4 h to mimic intestinal digestion. The results of GABA stability and the survival of probiotic cultures are shown in Fig. 2, as GABA content and viable cell count, respectively. No signi cant reduction of GABA was observed after 6 h of digestion. While gastrointestinal enzymes such as pancreatin and pepsin are available to perform hydrolysis in the simulated system, GABA is not digested because it is a non-protein amino acid, i.e. not a substrate for these enzymes. Instead, the structural integrity and stability of GABA is maintained through resistance towards acidic pH. Similarly, a study by Le et al. 31 revealed no signi cant reduction on GABA content in germinated soymilk after 2 h of simulated digestion at 37°C.
In contrast, a slow reduction in probiotic viability was detected over the course of hydrolysis, with descending values from 9.31 log CFU/g at 0 h to 8.49 log CFU/g at the end of 6 h digestion. A similar reduction was reported for L. acidophilus La-5 in fermented soy product 32 and B. animalis subsp. lactis in goat's milk ice cream 33 under the same assay conditions. Probiotics is only deemed bene cial and useful when they could tolerate harsh acidic condition in stomach and withstand further digestion in small intestine to reach large intestine in viable form, restoring gut microbial balance. While viable count reduced signi cantly after digestion in current study, the viability is not totally destroyed, but rather, maintained at 8.49 log CFU/g at the end of digestion. This is very likely due to a complementary effect from the acid-resistance nature of bacteria and H + ion dependent GABA production cycle. Coherently, Sanchart et al. 34 reported the survival of LAB at pH < 2.5 for at least 2 h, while Wang et al. 35 reported that GABA production involves the consumption of H + ion from extracellular environment, making it less acidic and favourable for probiotic survival. This study highlights the stability of UPMC90 and UPMC91 LAB strains and starter culture to maintain viable cell count after simulated digestion.

28-days Storage Stability Study
GABA content, viable cell count and pH The GABA content and probiotic viability during storage are vital to ensure good bio-functionalities and health bene ts to be delivered to consumers without jeopardising its organoleptic property (in terms of sourness measured as pH). The effect of 28-days refrigerated storage (2-4°C) on GABA content, viable cell count and pH of GABA-rich yogurt are illustrated in Fig. 3. A signi cant increase in GABA content was observed over 28-days storage compared to freshly-fermented yogurt on day 1 (59.00 mg/100g). GABA content reached the maximum of 113.95 mg/100g on day 21 and decreased to 83.65 mg/100g on day 28. Despite of this reduction, GABA content was still higher than the initial value, indicating that refrigerated storage up to 28 days is acceptable, but storage of 21 days is the optimum. The overall improvement of GABA content over 28 days storage period re ects the continuous formation of GABA by the GABA-producing LAB strains of UPMC90 and UPMC91 as well as starter culture throughout the study period. The increasing GABA content during storage is in agreement with previous nding 36  The viable cell count of GABA-rich yogurt was initially recorded at 9.68 log CFU/g on day 1 and peaked at 10.23 log CFU/g on day 14, followed by a sharp reduction to 9.17 and further to 9.06 log CFU/g on day 21 and 28, respectively. While the viability reduced over time, it is still in accordance to the minimum standard count of 6.00 log CFU/g as required for probiotic food recognition 37 , thus fostering the potential of GABA-rich yogurt as a probiotic food during 28 days of storage. An interesting observation was made on the reduced viable cell count on day 21 which recorded the highest GABA content. This is due to the fact that GABA is accumulated progressively over time and is rarely converted into other nal products by microbial reaction, as GABA is not a preferred substrate compared to other sugars in the fermentation medium, thus allowing the cumulative detection. In contrast, the reduced GABA content at the end of storage (day 28) was possibly due to structural instability in GABA which led to molecular selfdegradation, along with minimal occurrence of microbial/enzymatic reactions. Since 28 days of storage draws early sign of GABA instability, extending storage time longer than 28 days is not recommended. In terms of pH, a signi cant reduction was observed from pH 3.99 to 3.88 in GABA-rich yogurt, due to the production of lactic acid by the viable cells via lactose fermentation in the milk during refrigerated storage 38 . These pH range falls within the acceptable limit of pH 3.7-4.6 for commercial yogurt thus complying with the product speci cation requirement.
In standard yogurt, GABA was produced at a minimal level by starter culture (a mixture of S. thermophilus and L. delbrueckii ssp. bulgaricus), recording values from 9.02 mg/100g (day 1) to 17.16 mg/100g (day 28) that were signi cantly lower than that of GABA-rich yogurt. While the GABA-producing ability of these two strains has been acknowledged 39 , it depicted a low conversion rate. Coherently, Watanabe et al. 40 observed a poor GABA production (less than 5 mM) after 48 h of milk fermentation by the said starter culture. Therefore, GABA-rich yogurt was successfully proven to contain higher amount of GABA compared to standard yogurt at all times, during 28 days of storage at refrigerated temperature.
Water holding capacity and syneresis Water holding capacity (WHC) and syneresis directly re ect the coagulum strength of yogurt as a semisolid, gel-like food product and is related to the textural and sensorial properties (mouthfeel, eating experience) of a product 41 . Syneresis, a common phenomenon in yogurt, is considered unfavourable to consumers owing to the presence of exudate/ uid release from the food matrix. As shown in Fig. 4a, the WHC of GABA-rich and standard yogurt recorded no signi cant changes over 28 days of storage, except for increment on day 7, indicating high stability of yogurt samples over time. The result from syneresis study (Fig. 4b) showed GABA-rich yogurt exhibiting a minor decrease in syneresis values (11.70-15.03%) compared to standard yogurt (20.79-21.63%) during storage. Low syneresis values are mostly contributed by the acidic environment in both samples helps to enhance the gel network to resist syneresis during storage 42 . According to Lobato-Calleros et al. 43 and Nguyen et al. 44 , the increasing WHC and reducing syneresis were resulted from the effective water molecule entrapment in the protein network structure. However, the effect of glucose addition on WHC and syneresis is not statistically different between GABA-rich yogurt and standard yogurt at the day of storage with p-value>0.05.

H-NMR Metabolomics Analysis
From a molecular point of view, yogurt comprises of hundreds of biomolecules including proteins, lipids, sugars, amino acids, organic acids, fatty acids, minerals and volatile aroma compounds that contribute to the overall avour and taste pro le. In this study, both GABA-rich and standard yogurt were freeze-dried prior to nuclear magnetic resonance (NMR) spectroscopy analysis to prevent signal interference from water molecules 45 . It is known that each microorganism induces metabolite changes via different metabolic pathway during fermentation of food 46 . Therefore in this study, a metabolomics approach based on 1 H-NMR technique was used to compare the major metabolite pro le between freeze-dried GABA-rich yogurt (GY) and freeze-dried standard yogurt (SY), since both samples were fermented from milk by different microorganisms, of which SY contains only starter culture of S. thermophilus and L. delbrueckii ssp. bulgaricus while GY contains additional GABA-producing LAB strains of UPMC 90 and UPMC91. Fig. 5 depicts the 1 H-NMR spectra for a mixture of metabolites, consisting of amino acid, sugar and organic acid (lactic acid). A total of 16 and 13 compounds were detected in GY and SY, respectively. The metabolite pro les, observed as different spectrum pattern between GY and SY, is possibly due to the strain-speci c metabolic activities of GABA-producing LAB strains (UPMC90 and UPMC91). Similarly, a previous study reported that the free phenolic content varied during fermentation of whole-grain barley when different species of lactobacillus were used 47 . Table 1 tabulates the amino acid, sugar and organic acid content of GY and SY. Brie y, GY comprised of 7 amino acids including GABA, glutamine, alanine, histidine, proline, cysteine and valine while SY comprised of only 4 amino acids including GABA, alanine, histidine and choline. In cases where amino acid was present in both samples, GY showed higher concentration than SY except for alanine. Similarly, GY demonstrated higher GABA content (97.65 mg/100g) than SY (25.10 mg/100g). This portrays the successful role of glucose added extrinsically into the fermentation medium to enhance natural GABA production by UPMC90 and UPMC91 LAB strains. Also, the slight GABA value in SY indicated that GABA was produced naturally by starter culture without the presence of GABA-producing strains.
In terms of sugar, 7 types of sugar (glucose, lactose, lactulose, trehalose, arabinose, galactose and Nacetylglucosamine) were identi ed from both GY and SY samples. Higher amounts of glucose (311.85 mg/100g), trehalose (1174.00 mg/100g) and galactose (549.48 mg/100g) were observed in GY compared to SY. The high concentration of glucose was due to the external addition of this sugar into the fermentation medium as GABA enhancer, thus detected at higher value than SY. As for galactose, it was excreted into the medium when the microorganisms (starter culture and GABA-producing LAB strains) consumed the glucose moiety of lactose in the milk 48 , leaving behind galactose residue. The increased level of galactose in GY indicated high consumption rate of lactose as the preferred substrate among starter culture co-inoculated with UPMC90 and UPMC91 LAB strains. Similar to our ndings, higher galactose content was obtained when co-culturing L. plantarum WCFS1 with S. thermophilus and L. delbrueckii ssp. Bulgaricus compared to that without L. plantarum WCFS1 49 . Lower amount of lactose correlates to the lower amount of lactic acid in GY. This could be explained by the heterofermentative metabolism of UPMC90 and UPMC91 strains in GY, which favour lactose utilisation for energy production while generating metabolites other than lactic acid, thus lowering the amount of lactic acid.
In contrast, starter culture (S. thermophilus and L. delbrueckii ssp. bulgaricus) in SY produced higher lactic acid due to homofermentative metabolism that generates lactic acid as the main end-product during fermentation 50 . While yogurt is widely recognised to be suitable for lactose-intolerant individuals, the signi cantly lower amount of lactose in GY (346.41 mg/100g), as compared to SY (1025.00 mg/100g), provides additional bene t to patients suffering from severe lactose intolerant symptoms. In short, the metabolites were produced as a result of major structural alteration of milk components through two biochemical pathways from the microorganisms: (i) glycolysis where carbohydrate was converted into lactic acid or other metabolites, and (ii) proteolysis where casein was hydrolysed into peptide or free amino acid 51 .

Conclusion
Current study represents a pioneer that reports the effect of simple sugars and prebiotics on enhancing natural GABA production in a food system (yogurt) along with metabolomics pro ling of the generated biomolecules. Simple sugars induced higher GABA production compared to prebiotics through natural fermentation by two novel, self-cloned L. plantarum Taj-Apis362 strains (UPMC 90 and UPMC91) in the yogurt. In particular, glucose served as a superior choice over PLP by successfully enhanced GABA production at a very low concentration of glutamate substrate (11.5 mM). Simulated gastrointestinal study and storage study coherently revealed good stability of GABA and viable cell count under gastrointestinal condition as well as refrigerated storage up to 28 days, meeting the minimum requirement of 6.00 log CFU/g for recognition as probiotic food. The addition of glucose also does not affect the water holding capacity and syneresis values of GABA-rich yogurt as there was no signi cant difference detected when compared with standard yogurt. 1 H-NMR analysis revealed different metabolomics pro le of GABA-rich yogurt and standard yogurt, detecting 16 and 13 compounds (amino acid, sugar and organic acid), respectively. This study successfully mitigates the over-use of glutamate substrate and omits the use of expensive PLP cofactor in the production of GABA-rich yogurt, offering an economical approach to produce a probiotic-rich, functional dairy food with prospective stress management and cardiovascular disease prevention properties.

Materials
Non-fat skim milk powder (Sunlac® brand) and pasteurized fresh milk (Goodday® brand) were locally purchased. Food grade commercial prebiotics (with 90-95% purity), inulin and galactooligosaccharides (GOS) were purchased from CK Chemical Sdn Bhd and fructooligosaccharides (FOS) was purchased from Green nite Sdn Bhd. MRS agar and MRS broth were obtained from HiMedia Laboratories Pvt. Ltd.
(Mumbai, India). Glutamate, GABA standard and triethylamine were obtained from Merck KGaA (Darmstadt, Germany). Methanol-d 4 , deuterium oxide (D 2 O) and sodium deuteroxide (NaOD) used for NMR analysis were purchased from Cambridge Isotope Laboratories, Inc. (Tewksbury, MA, USA). All other chemicals used were of analytical or HPLC grade.
Preparation of Starter Culture and GABA-producing LAB strains Two L. plantarum Taj-Apis362 strains possessing high intracellular GAD activity (UPMC90) and high extracellular GAD activity (UPMC91) were obtained from the culture collection of Institute of Bioscience, Universiti Putra Malaysia and routinely stored in sterile MRS broth at -80°C as stock culture. These strains were characterized from previous work, where the wild-type L. plantarum Taj-Apis362 was previously isolated from the stomach of honeybee Apis dorsata, and used as a host for GAD gene overexpression to produce UPMC90 and UPMC91 strains 11 . All procedures involving the use of L. plantarum Taj-Apis strains received approval from National Board of Biosafety, Ministry of Natural Resources and Environment, Malaysia (approval no. JBK [S]-602-1/2/207). Commercial starter culture (Lactina® brand) containing S. thermophilus and L. delbrueckii ssp. bulgaricus was obtained from YogurtBio (So a, Bulgaria). Reconstituted skim milk was prepared by mixing commercial pasteurized fresh milk and skimmed milk powder to reach 16% of nonfat dry matter. It was then submitted to thermal treatment at 80-85°C for 30 min 44 , cooled to 4°C in an iced bath, poured into 250-mL screw capped Schott bottles and stored at 4°C for 24 h before used. At the beginning of each fermentation cycle, i.e. production of new batch of yogurt sample, starter culture and LAB strains were prepared freshly from the stock. Starter culture was inoculated in a sterilised reconstituted skim milk prepared previously and incubated at 42°C for about 6 h until reaching pH 4.5-4.6. On the other hand, LAB strains were streaked for single colony isolation on MRS agar, then transferred to 10 mL MRS broth, incubated for 18 h at 37°C to allow cell growth, and further transferred to sterilised reconstituted skim milk for sub-culturing for 22-24 h. At the end of incubation, the coagulated milk was employed as inoculum for yogurt production.

Addition of Sugars and Prebiotics into Yogurt
Eight samples of yogurt were prepared: negative control (yogurt with glutamate only), positive control (yogurt with glutamate + 50 µM of PLP cofactor), and yogurt with glutamate + 2% (w/v) of glucose, sucrose, fructose, inulin, FOS and GOS, respectively. Brie y, the yogurt samples were prepared by coinoculating starter culture and GABA-producing LAB strains (having viable count of 10 6 CFU/g) simultaneously at a ratio of 2:1 w/w, into fresh sterilised reconstituted skim milk. A total of 11.5 mM of glutamate was added and fermentation was allowed for 7.25 h at 39°C, which were the optimum conditions obtained from previous work 19 . Upon completion, samples were rapidly cooled in an ice bath to stop further fermentation and stored at 2-4°C until further analysis.
Determination of GABA Content GABA and glutamate were determined following the method previously described by 11 using HPLC system (Shimadzu LC 20AT, Shimadzu Corporation, Kyoto, Japan) equipped with oven (model CT0-10ASVP), pump system and PDA detector (model SPD-M20A). Chromolith® RP-18 endcapped separation column (100 mm length × 4.6 mm internal diameter, Merck KGaA, Darmstadt, Germany) was used for this analysis. Yogurt sample was centrifuged at 10,000 x g for 15 min at 4°C and 10 μL of the supernatant was placed into small durham tube and evaporated under vacuum for 40 min. Then, the dried supernatant was dissolved in 20 μL of a mixture of ethanol/water/triethylamine solution at a ratio of 2:2:1 and vacuum-evaporated for another 40 min, followed by addition of 30 μL of a mixture of ethanol/water/triethylamine/phenylisothiocyanate solution at a ratio of 7:1:1:1 and left for 20 min at room temperature to allow phenylisothiocyanate-GABA formation. The sample was vacuum-evaporated again for 40 min to remove excess reactant.
The derivatized sample was then diluted and subjected to HPLC analysis. Mobile phase A was prepared by dissolving 8.205 g of sodium acetate, 0.5 mL of trimethylamine and 0.7 mL of acetic acid in 1000 mL of deionized water, then the pH was adjusted to 5.8 using 0.1 M sodium hydroxide. Meanwhile, mobile phase B was prepared by mixing acetonitrile with deionized water at a ratio of 60:40 (v/v). Both mobile phases were ltered through a 0.45 μm membrane lter. Sample was injected at 5 µL and eluted at a ow rate of 0.6 mL/min using isocratic elution of 80% mobile phase A + 20% mobile phase B. Compound detection was performed using a diode array detector at λ = 254 nm. The GABA and glutamate contents were calculated by comparing the sample peak area with GABA standard and glutamate standard, respectively.

Viable Cell Count
Bacterial enumeration was performed using the pour plate method. Firstly, 1.0 g gram of yogurt sample was diluted with 9.0 mL of sterile peptone water. Subsequently, a 10-fold dilution was made using peptone water, and 0.1 mL of the diluted sample was spread on MRS agar and cultured at 37 C for 48 h to allow cell growth. The colonies appearing on the plates were then counted, multiplied by dilution factor, and expressed as log colony forming unit per g (log CFU/g).

Gastrointestinal Stability Study (Simulated Digestion)
From six yogurt samples added with different sugars and probiotics, sample showing the highest GABA content and viable cell count was selected for further product performance evaluation and characterisation as follows: Gastrointestinal stability study, 28-days storage stability study and metabolomics pro ling. Simulated digestion was performed following the method described by 52 . Two solutions were prepared, namely simulated-gastric-uid (SGF) and simulated-intestinal-uid (SIF). The SGF was prepared by mixing 20 mg/mL of pepsin, 350 µL of concentrated HCl and 0.1 g of NaCl in deionized water to a total volume of 50 mL and the pH was adjusted to 1.2. Next, 1 mL of the SGF solution was added to 3 mL of yogurt and incubated at 37°C in water bath shaker for 2 h. Meanwhile, SIF solution was prepared by mixing 34 mg/mL of KH 2 PO 4 , 3.85 mL of NaOH (200 mM) and 0.5 g of pancreatin in deionized water to a nal volume of 50 mL and the pH was adjusted to 6.8. Then, 1 mL of the SIF solution was added to the reaction mixture and re-incubated for 4 h under the same condition.
Aliquots of 1 mL were taken at 0, 2, 4 and 6 h and boiled at 100 C for 10 min to inactivate enzymes and stored at -20°C for GABA content. For viable cell count, aliquots of 1 mL were also taken at 0, 2, 4 and 6 h and rapidly cooled before storing at -20°C.

pH determination
The pH value of yogurt samples was measured using a pH meter (model S20 SevenEasy TM , Mettler-Toledo GmbH, Columbus, OH, USA).

Water holding capacity and syneresis
The water-holding capacity (WHC) of yogurt was determined according to the modi ed procedure described by 53 . Yogurt sample of 10 g (W 1 ) was centrifuged at 5000 g for 10 min at 4°C. The supernatant was collected and weighed (W 2 ). WHC (%) was calculated as follows. WHC = W 1 -W 2 /W 1 X 100% (1) Syneresis was determined according to the method from 54 . Brie y, 10 g of yogurt sample (W 1 ) was centrifuged at 700 g for 10 min at 4°C. The supernatant was collected and weighed (W 2 ). Degree of syneresis (%) was calculated as follows.
Degree of syneresis = W 2 /W 1 X 100 Metabolomics Pro ling ( 1 H-NMR Analysis) Yogurt samples were freeze dried and subjected to 1 H-NMR analysis as described by 55 . A total of 10 mg of freeze-dried yogurt was mixed with 0.375 mL of CH 3 OH-d 4 and 0.375 mL of KH 2 PO 4 buffer in D 2 O, containing 0.1% trimethylsilyl propionate as internal standard. The pH was adjusted to 6 with NaOD. The mixture was vortexed for 1 min, sonicated in an ultra-sonicator at 30 C for 15 min and centrifuged at 13,000 rpm for l0 min. Supernatant aliquot of 600 μL was transferred to NMR tube for 1 H-NMR analysis. Spectra were recorded at 26 C on a spectrometer (model UNITY INOVA 500, Agilent Technologies Inc., Santa Clara, CA, USA) using a frequency of 500 MHz. Tetramethylsilane was used as an internal standard. The spectra were automatically phased and bucketed with standard bins of δ 0.05 ranging from region δ 0.50 to 10.00. The metabolites were identi ed using Chenomx software version 8.5 (Chenomx Inc., Edmonton, Canada). The residual methanol region (δ 3.28 to 3.33) and water region (δ 4.70 to 4.96) were excluded from the analysis. Two-dimensional 1 H-1 H J-resolved and Heteronuclear Multiple-Bond Correlation (HMBC) were employed for the metabolites identi cation. Six replicates were examined for each yogurt sample.

Statistical Analysis
Analysis of variance (ANOVA) followed by Duncan's test and 2-sample t-test were used to evaluate means at signi cant difference of p<0.05 using Minitab software version 16 (Minitab Inc., State College, PA, USA). All values were reported as means ± standard deviation from at least triplicate determinations.