The three organisms at the heart of the study were not chosen at random. Lactobacillus paracasei 7060 was originally isolated from the sap of Nypa fruticans, the salt-tolerant nipa palm that grows along Philippine coastlines. Limosilactobacillus fermentum PJG11 came from Dioscorea luzonensis Schauer, an indigenous tuber known locally as kamangeg, while Lentilactobacillus hilgardii DSB22 was recovered from Dendrocalamus merrillianus Elmer, the endemic bayog bamboo. These plant-associated niches are considered underexplored microbial reservoirs, and the researchers argued that strains adapted to such ecologically distinct environments may harbor metabolic capabilities that industrially familiar laboratory strains simply do not possess. Previous work had already sequenced the genomes of the three isolates, but genome sequences alone reveal potential rather than actual chemical output.

To capture that actual output, the team grew each bacterium in de Man, Rogosa and Sharpe broth, a rich laboratory medium, and then removed the cells by centrifugation and fine filtration to obtain cell-free supernatants. Because the growth medium itself is loaded with sugars, salts and peptides that could mask microbially derived compounds, the researchers extracted the semi-polar metabolites with ethyl acetate and, crucially, processed uninoculated medium blanks through exactly the same workflow. Any signal that appeared in both the cultures and the blank could not be attributed to bacterial activity. The extracts were then separated on an ultra-high-performance liquid chromatography system coupled to a quadrupole time-of-flight mass spectrometer, an instrument combination capable of measuring the precise masses and fragmentation patterns of hundreds of small molecules in a single run.

After rigorous data processing, 133 features survived blank subtraction and formed the basis of the comparative analysis. Unsupervised principal component analysis accounted for 58.3 percent of the total variance across the dataset and revealed clear spatial separation: each of the three bacterial groups clustered apart from the uninoculated medium and, importantly, apart from each other. This was direct evidence that the three indigenous strains exhibit genuinely distinct extracellular metabolomes. Quality control samples, pooled from all experimental extracts and injected periodically throughout the analytical sequence, clustered tightly, confirming that the observed separations reflected real biology rather than instrumental drift. One strain, L. fermentum PJG11, showed greater variability among its biological replicates, hinting at a more variable metabolic output under laboratory conditions.

Differential abundance analysis, using volcano plots that combined statistical significance testing with an eightfold change threshold, sharpened the picture. Across all three strains, five features were consistently enriched relative to the medium: L-glutamic acid, fumaric acid, indole-3-lactic acid, perlolyrine and pyroglutamylisoleucine. The first three are well documented among the extracellular products of probiotic bacteria and yeasts, but perlolyrine, a furan-containing beta-carboline alkaloid previously reported mainly in plants, fungi and processed foods, and the pyroglutamyl peptide PyroGlu-Ile stood out as far less familiar microbial associates. Notably, indole-3-lactic acid, perlolyrine and PyroGlu-Ile were completely absent from the uninoculated medium and appeared only after bacterial growth, strongly suggesting that the cultures themselves were responsible for their presence. A heatmap analysis further highlighted a cluster of diketopiperazines, cyclic dipeptides common in fermented foods, several of which were enriched in the bacterial cultures even though related features arose in the medium through heat-induced cyclization during sterilization.

Supervised statistical models then pushed the discrimination further. A partial least squares-discriminant analysis, validated by 100-fold permutation testing, and a Random Forest model built from 500 decision trees, which achieved a 6.7 percent out-of-bag error rate, both pulled L. paracasei 7060 apart from the other two strains, while L. hilgardii DSB22 and L. fermentum PJG11 partially overlapped. The researchers connected that overlap to shared fermentative physiology: both are obligately heterofermentative lactic acid bacteria that rely on the phosphogluconate pathway for hexose fermentation, whereas L. paracasei 7060 is facultatively heterofermentative and can flexibly deploy both the Embden–Meyerhof–Parnas and phosphoketolase routes depending on the substrate. Variable importance in projection scores identified the most discriminatory molecules: perlolyrine and 1-acetyl-beta-carboline, both tentatively linked to beta-carboline alkaloids, distinguished L. paracasei 7060, leucyl-proline characterized L. fermentum PJG11, and indole-3-lactic acid marked L. hilgardii DSB22.

The metabolomic separation tracked measurable physiological differences. L. paracasei 7060 grew rapidly early on, drove the pH down furthest and reached the highest titratable acidity, 2.42 percent, at 20 hours, while the other two strains acidified more gradually and reached lower acidity. At the 48-hour sampling point used for chemical profiling, the three cultures were in different growth states: L. paracasei 7060 was already in decline, L. fermentum PJG11 sat near its plateau, and L. hilgardii DSB22 was approaching its maximum viable cell density. The authors were careful to note a limitation: because the study did not independently separate species identity from plant origin, the relative contributions of each factor to the observed chemical differences cannot be determined, although plant-associated bacteria are known to develop niche-specific carbohydrate utilization and stress tolerance traits.

To give the chemical data biological context, the team turned to functional genomics, focusing on L. fermentum PJG11. Its genome yielded 756 uniquely annotated KEGG Orthology assignments, with carbohydrate and amino acid metabolism the most heavily represented functional categories. A flux balance analysis model, built on the KBase platform, showed positive modeled flux toward fumarate formation via aspartate ammonia-lyase and toward glutamate production through a gap-filled transamination reaction, consistent with the enrichment of fumaric and L-glutamic acid in the cultures. The genome also carried plausible enzymatic machinery for the other signature compounds: an X-Pro dipeptidyl peptidase gene, PepX, could generate the leucyl-proline dipeptide from imported oligopeptides, while aminotransferase candidates PatA and HisC, together with lactate dehydrogenase-like proteins, could plausibly convert L-tryptophan through indole-3-pyruvate into indole-3-lactic acid, even though the canonical genes for that pathway were absent.

The beta-carboline story proved more cautionary and scientifically nuanced. Although recent literature suggests that lactic acid bacteria can produce beta-carbolines de novo, the feature matching 1-acetyl-beta-carboline appeared in both inoculated cultures and the uninoculated medium, echoing earlier reports that this compound can form abiotically during medium sterilization, so its increased abundance in cultures cannot be unequivocally credited to bacterial biosynthesis. Perlolyrine, by contrast, was detected exclusively in bacterial cultures, suggesting an association with growth, yet its precise origin remains unresolved because the enzymatic routes to this alkaloid are largely unexplored. The researchers sketched a plausible supply route: bacterial acidification and glucose metabolism could generate reactive carbonyl intermediates such as methylglyoxal and 3-deoxyglucosone, which react with L-tryptophan to build these alkaloids. The flux model supported glucose routing through glycolysis, and the genome encoded triosephosphate isomerase, but the methylglyoxal synthase gene mgsA was absent, leaving nonenzymatic degradation of triose phosphates as the tentative source.

Beyond its immediate findings, the study expands the exometabolomic record for a group of organisms that has long been dominated by industrial strains from familiar food plants. By showing that three indigenous Philippine isolates, each drawn from a different endemic or native host plant, produce measurably distinct extracellular chemistries, the work reinforces a growing consensus in microbiology that metabolomic traits cannot be generalized across bacterial taxa and must be characterized strain by strain. It also delivers a shortlist of priorities for future research: perlolyrine and pyroglutamylisoleucine, both only putatively annotated here, now stand as concrete targets for targeted screening, confirmatory identification and bioactivity testing within the fast-growing field of postbiotics, where bacterial fermentation products, rather than the bacteria themselves, are harnessed for food preservation, health promotion and functional ingredient development. The tropical and endemic flora of the Philippines, the study suggests, may still hold many more such chemical surprises.

Subject of Research: Exometabolomic profiling of indigenous Philippine plant-derived lactic acid bacteria

Article Title: Comparative exometabolomic insights into indigenous Philippine plant-derived lactic acid bacteria

Article References: Diez, K. J. T., Bucao, X. E. N., Rafael, C. A. R., Batuyong-Lista, M. A. R., Valerozo, J. A., Ong, J. P., Bucao, D. S., Agrupis, S. C., & Ramil, R. J. D. (2026). Comparative exometabolomic insights into indigenous Philippine plant-derived lactic acid bacteria. Journal of Agriculture and Food Research, Article 103367. https://doi.org/10.1016/j.jafr.2026.103367

Image Credits: AI Generated

DOI: 10.1016/j.jafr.2026.103367

Keywords: lactic acid bacteria, exometabolomics, Philippines, postbiotics, metabolomics, Nypa fruticans, Dioscorea luzonensis, Dendrocalamus merrillianus, beta-carbolines, indole-3-lactic acid, UHPLC-Q-TOF-MS, functional genomics