Untargeted mass spectrometry-based metabolomics approach unveils biochemical changes in compound probiotic fermented milk during fermentation

Fermentation is a metabolic process in which organic matter is decomposed by enzymes. Fermented milk is a major fermented food, recognized as healthy and an effective carrier of probiotics. Probiotic fermented milk has been associated with multiple health benefits (e.g., cholesterol lowering, immune modulation, gut health, cancer prevention, mitigation of cognitive impairment) attributable to functional components such as peptides, polysaccharides, fatty acids, organic acids, vitamins, and γ-aminobutyric acid (GABA). Lacticaseibacillus paracasei is widely used in probiotic foods with reported effects on gastrointestinal health, immunity, and oral health, and many strains have good fermentation traits. Bifidobacterium adolescentis is an important member of the human gut microbiota linked to health (e.g., body weight maintenance, prevention of constipation), but it is difficult to use in fermentation due to low viability; co-fermentation with strains possessing good fermentation performance can improve its viability. Previous work has studied metabolomic changes in probiotic fermented milk, but typically in combination with traditional starters (Streptococcus thermophilus and Lactobacillus delbrueckii subsp. bulgaricus), making it difficult to distinguish probiotic-specific effects. Moreover, ripening changes may occur due to ongoing microbial activity post-fermentation, yet prior studies often focus on ripening without monitoring metabolite dynamics during fermentation. This study co-fermented milk with Lacticaseibacillus paracasei PC-01 (from yak milk) and Bifidobacterium adolescentis B8589 (from an infant gut), previously shown to increase GABA and short-chain fatty acids versus PC-01 alone. The objective was to investigate time-course metabolomic changes during co-fermentation and ripening, alongside probiotic viable counts, to elucidate probiotic-specific metabolic alterations in a milk matrix.

The authors note that most metabolomics studies of fermented milk employ traditional starter cultures together with probiotics, hindering attribution of metabolite changes specifically to probiotic activity. Traditional starters (Streptococcus thermophilus and Lactobacillus delbrueckii subsp. bulgaricus) can confound interpretation of probiotic-specific metabolism. Prior studies often emphasize ripening-stage changes rather than continuous monitoring throughout fermentation, limiting understanding of dynamic metabolic transitions that relate to product quality and stability. The study builds on evidence that co-fermenting B. adolescentis with a robust fermenter (e.g., L. paracasei PC-01) can enhance production of functional metabolites such as GABA and short-chain fatty acids.

Study design: Pasteurized milk was co-fermented anaerobically at 37 °C by two probiotic strains, Lacticaseibacillus paracasei PC-01 and Bifidobacterium adolescentis B8589, until the pH reached 3.8 at 72 h. Samples were collected every 12 h; metabolomics analyses were performed at 0, 36, 60, and 72 h. Probiotic viability was monitored by flow cytometry. Fermented milk production: Skim milk powder (NZMP) with 2% glucose was dissolved in water (50 °C, 30 min), homogenized (65 °C, 20 MPa), and pasteurized (95 °C, 5 min), then cooled to 20 °C. Initial inocula were adjusted to 1×10^7 CFU/mL for PC-01 and 1×10^9 CFU/mL for B8589. Fermentation proceeded anaerobically at 37 °C to pH 3.8 (72 h), after which samples were stored at −80 °C. Viable cell enumeration: Samples were diluted to 10^6–10^7 CFU/mL. SYTO 9 and propidium iodide (PI) staining (single and double) were employed. After staining (15 min, dark, room temperature), double-stained samples were mixed with absolute counting microspheres and analyzed on a MoFlo Astrios EQ flow cytometer (argon-ion laser, 488 nm excitation; emission >630 nm). Gating strategies quantified SYTO 9-positive/PI-negative viable cells; absolute counts were derived using microsphere standards. Untargeted metabolomics: 100 µL sample was extracted with 400 µL acetonitrile:methanol (1:1) containing isotopically labeled internal standards. After sonication on ice (10 min), standing at −4 °C (1 h), and centrifugation (4 °C, 12,000 rpm, 15 min), supernatants were filtered for UPLC-QE-MS analysis. Chromatography used an Acquity UPLC BEH Amide column (2.1×100 mm, 1.7 µm) with mobile phases: A (25 mM ammonium acetate and 25 mM ammonia in water) and B (acetonitrile). Gradient: 0–0.5 min 95% B; 0.5–7 min 95–65% B; 7–8 min 65–40% B; 8–9 min 40% B; 9–9.1 min 40–95% B; 9.1–12 min 95% B. Tray 4 °C; injection 2 µL. MS acquisition used a Q Exactive HF-X in IDA mode (Xcalibur): ESI source, sheath gas 3.35 L/min, aux gas 16.8 L/min, capillary 350 °C, m/z 70–1050, cycle 760 ms, full MS resolution 60,000, MS/MS resolution 7,500, NCE 10/30/60, spray voltage +3.6 kV/−3.2 kV. QC samples were pooled aliquots of all samples to monitor system stability. Data processing and statistics: Raw files were converted to mzXML (ProteoWizard). A custom R pipeline (XCMS kernel) performed peak picking, extraction, alignment, and integration; metabolites were annotated against an in-house MS2 database (BiotreeDB). PCA and OPLS-DA (MetaboAnalyst 5.0) visualized and modeled differences; model quality metrics R^2Y and Q^2 were reported. Differential metabolites were selected via volcano plots with thresholds fold change >2 or <0.5 and P<0.05 (ANOVA). Pathway annotation and enrichment were conducted using KEGG; significance assessed by Wilcoxon rank-sum test.

- Probiotic viability: Viable counts peaked at 36 h (1.57×10^10 CFU/mL), significantly higher than at 0 h (8.49×10^7 CFU/mL; P<0.001), decreased at 60 h (8.87×10^9 CFU/mL; P<0.001 vs 36 h), and rose again by 72 h (1.57×10^10 CFU/mL; P<0.001 vs 60 h). - Global metabolome dynamics: PCA showed clear time-based clustering of samples at 0, 36, 60, and 72 h with tight QC clustering, indicating stable analytical performance. OPLS-DA models demonstrated strong group separation with high model performance: 0 vs 36 h (R^2Y=0.999, Q^2=0.997), 36 vs 60 h (R^2Y=0.992, Q^2=0.983), 60 vs 72 h (R^2Y=0.994, Q^2=0.988). - Differential metabolites: 0–36 h: 304 features (151 up, 153 down; 244 identified) across classes including lipids (66), organic acids/derivatives (59), organic oxygen compounds (43), organoheterocycles (31), phenylpropanoids/polyketides (13), benzenoids (15), and others. 36–60 h: 110 features (43 up, 67 down; 79 identified) with organic oxygen compounds (21), organic acids/derivatives (20), organoheterocycles (17), lipids (9), others (12). 60–72 h: 36 features (24 up, 12 down; 27 identified) including lipids (7), organic oxygen compounds (9), organic acids/derivatives (5), others (6). - Pathway enrichment: 0–36 h: 25 pathways for organic acids/derivatives (amino acid and organic acid metabolism, including TCA cycle, alanine/aspartate/glutamate, D-glutamine/D-glutamate, pyruvate metabolism), 10 lipid-related pathways (fatty acid biosynthesis/elongation/degradation, glycerolipid, ether lipid, alpha-linolenic, arachidonic acid, primary bile acid), and 10 organic oxygen compound pathways (carbohydrate metabolism: amino sugar, pentose phosphate, starch/sucrose, pentose/glucuronate, ascorbate/aldarate, purine, nicotinate/nicotinamide, inositol phosphate, phenylpropanoid). 36–60 h: 17 organic acid/derivative pathways (including TCA, amino acid metabolism), 3 lipid pathways (glycerophospholipid, glycerolipid, ether lipid), 7 organic oxygen compound pathways (carbohydrates). 60–72 h: 2 organic acid/derivative pathways, 2 lipid pathways (biosynthesis of unsaturated fatty acids, primary bile acid), 6 organic oxygen compound pathways (pentose phosphate, amino sugar, sphingolipid, phosphatidylinositol signaling, purine metabolism). - Key stage-specific metabolites: Nine biomarkers linked to TCA cycle, glutamate metabolism, and fatty acid metabolism (succinic acid, pyruvic acid, L-glutamic acid, fumaric acid, GABA, capric acid, oleic acid, palmitic acid, stearic acid). Pyruvate and succinic acid increased notably by 36 h; fumaric acid decreased. GABA increased continuously, peaking at 72 h. Oleic acid decreased after 36 h, while stearic acid increased; palmitic acid increased. Capric acid increased by 36 h and remained elevated through 72 h. - Overall temporal pattern: Largest metabolomic changes occurred early (0–36 h), with milder differences during 36–60 h and 60–72 h (ripening). Functional biomolecules (pyruvate, GABA, capric acid) increased by the end of fermentation, potentially enhancing nutritional and functional properties.

The study delineates probiotic-specific metabolic dynamics in milk during co-fermentation by L. paracasei PC-01 and B. adolescentis B8589. The marked early shift (0–36 h) in the metabolome coincides with rapid bacterial growth, supported by increased glycolytic flux yielding pyruvate, which also feeds acetyl-CoA and the TCA cycle. Accumulation of succinic acid alongside decreasing fumaric acid suggests altered TCA branch activity and possible consumption/reduction of fumarate or rapid conversion to malate. Amino acid metabolism closely interfaces with the TCA cycle; conversion of L-glutamate to GABA via glutamate decarboxylase is facilitated by fermentation-induced acidification. Continuous GABA accumulation correlates with active probiotic viability and may further support bacterial growth while improving sensory properties by reducing glutamate’s salty note. Lipid metabolism changes reflect interactions with central metabolism via acetyl-CoA. The observed decrease in oleic acid with concomitant increase in stearic acid is consistent with hydrogenation during fermentation; increases in palmitic and capric acids also occurred. These fatty acid shifts, together with increases in organic acids and amino acid derivatives, contribute to flavor, texture, and potential health-relevant properties of the fermented milk. Collectively, enriched pathways (TCA, amino acid, fatty acid, and carbohydrate metabolism) explain the time-resolved biochemical remodeling and address the research aim of characterizing probiotic-specific fermentative changes and their functional implications.

Time-course UPLC-QE-MS metabolomics combined with flow cytometry revealed substantial, stage-dependent biochemical remodeling in probiotic co-fermented milk. The most pronounced changes occurred between 0 and 36 h, with comparatively modest shifts during 36–60 h and ripening (60–72 h). Differential metabolites predominantly comprised organic acids, amino acids, and fatty acids, implicating the TCA cycle, glutamate/GABA metabolism, and fatty acid metabolism. Nine time point-specific biomarkers (succinic acid, pyruvic acid, L-glutamic acid, fumaric acid, GABA, capric acid, oleic acid, palmitic acid, stearic acid) underpinned these changes. Functionally relevant increases in pyruvate, GABA, and capric acid by the end of fermentation may enhance nutritional quality and probiotic properties. The study provides detailed insight into probiotic metabolism in a milk matrix and a basis for optimizing fermentation for targeted functional metabolites. Future work should elucidate unreported interactions among key metabolites and verify health effects of altered fatty acid profiles and other metabolites in vivo.

The authors note that some interactions among identified metabolites and associated pathways have not been previously reported and merit further investigation. They also state that more studies are needed to confirm the health effects of fatty acid changes and other metabolites observed; the study did not directly assess health outcomes.