Reading the Living Recipe: Metatranscriptomics Maps Gene Expression in Traditional Fermented Foods

A sweeping review of metatranscriptomic research on traditional fermented foods has mapped how scientists are using RNA sequencing to eavesdrop on the microbial activity that shapes kimchi, cheese, sourdough, soybean paste, vinegar, and Chinese Baijiu. The synthesis surveyed publications indexed in the Web of Science between 2009 and 2 August 2026, retrieving 107 records that addressed metatranscriptomic analysis of fermented foods, beverages, or food-fermentation microbiomes. A keyword co-occurrence network built from 42 recurring terms revealed three thematic clusters: one centered on microbiota composition and product quality, another on microbial gene expression and enzyme function, and a third on metabolic pathways underlying flavor compounds and organic acids. The breadth of that map underscores both the growing appeal of transcriptomic approaches and the interpretive caution the authors insist must accompany every conclusion drawn from RNA data alone.

Fermentation is among humanity’s oldest food-processing techniques, valued not only for preservation but for the distinctive flavors and nutritional enhancements it imparts. Traditional fermented foods, whether prepared in household kitchens from locally sourced ingredients or manufactured through industrial-scale processes, harbor diverse microbial communities that researchers have described as part of humanity’s extended microbiome. These communities can serve as dietary sources of live microorganisms, although their viability, safety, and effects on the host depend on the specific product and strain involved. Growing calls to include fermented foods in dietary guidelines reflect this recognition, yet the microbial communities within them vary dramatically with geography, raw materials, processing methods, and maintenance practices, complicating efforts to identify reproducible community patterns or to define the core microbiota responsible for consistent product quality.

Metatranscriptomics occupies a distinctive position among culture-independent meta-omics approaches. Where metagenomics profiles community DNA to characterize which organisms are present and what genes they encode, metatranscriptomics profiles community RNA to reveal which genes are actually being transcribed under the sampled conditions. This distinction matters enormously for fermentation research, because a microorganism may carry the genetic blueprint for a pathway without ever activating it. Transcript abundance, however, does not directly establish protein abundance, enzyme activity, metabolic flux, or metabolite production. The review’s authors apply a strict evidence hierarchy throughout, distinguishing experimentally validated mechanisms from associations inferred solely from transcript annotation or pathway reconstruction, and reserving the term active microbiota for transcriptional signals unless corroborated by proteomic, metabolomic, or culture-based evidence.

The technical workflow begins with sample collection, where microbial RNA degrades rapidly and must be stabilized before extraction. Immediate snap-freezing in liquid nitrogen remains the most widely recommended approach for fermented-food samples, though commercial RNA-stabilizing agents permit short-term room-temperature storage with the caveat that they may not fully prevent degradation or transcriptional shifts. RNA can be extracted directly from the food matrix or after separation of microbial cells, each strategy carrying trade-offs. Direct extraction preserves native transcript profiles but co-extracts abundant non-target substrate transcripts from plant, animal, or fungal matrices, reducing the proportion and effective sequencing depth of microbial mRNA. Cell separation improves yield but may perturb community composition and induce stress-related transcriptional changes, meaning that the choice between approaches can itself shape the biological signal a study reports.

Complex food matrices introduce formidable analytical obstacles. Fermented foods often contain lipids, polyphenols, starch, and salts that interfere with RNA extraction, necessitating matrix-specific pretreatments such as sodium chloride to remove starch from fermented grains or trisodium citrate to disperse the cheese matrix. Ribosomal RNA typically accounts for more than 80 percent of total RNA and must be depleted before sequencing, yet depletion efficiency and taxonomic bias depend on community composition and probe compatibility. The scale of matrix-dependent variation is striking: RNA yields from 32 commercial cheeses ranged from 1 to 363 micrograms per gram of cheese, with lower RNA integrity reported for pressed-cooked varieties. Such pronounced differences indicate that the feasibility of any given workflow, particularly full-length long-read sequencing approaches, must be evaluated separately for each food matrix rather than assumed to transfer across systems.

Among fermented fruits and vegetables, metatranscriptomic studies have linked flavor formation to carbohydrate fermentation and acidification, phenolic-acid and volatile metabolism, and sulfur metabolism. In kimchi, transcripts related to carbohydrate metabolism and lactic acid fermentation were detected in major lactic acid bacteria, with late-stage expression of heterolactic fermentation genes reported in Weissella koreensis. During Sichuan radish paocai fermentation, transcripts encoding phenylpyruvate decarboxylase, oxaloacetate decarboxylase, and alcohol dehydrogenase were associated with the accumulation of flavor-related compounds, making these enzymes candidates for functional testing. In a separate validation study, three candidate phenolic acid decarboxylases were expressed in Escherichia coli, and the enzyme from Lactobacillus vermolensis showed the highest catalytic activity, broad pH tolerance, and good thermal stability, with preliminary application enhancing vinylphenol aroma. Transcripts encoding nitrite reductases nrfA and nirB, associated with Enterobacteriaceae during paocai fermentation, highlight food-safety concerns that warrant further investigation into nitrite dynamics.

Fermented soybean products present a different set of challenges centered on biogenic amines, nitrogenous compounds formed through bacterial decarboxylation of amino acids that can pose toxicological risks at excessive intake levels. Integrated metagenomic and metatranscriptomic analysis of traditional Korean ganjang fermentation identified 16 candidate biogenic-amine decarboxylase genes, including ornithine, lysine, histidine, and tyrosine decarboxylases distributed across Bacillus, Halomonas, Staphylococcus, Tetragenococcus, and Lactobacillus in stage-specific patterns. These results implicate the corresponding taxa as candidate contributors to biogenic-amine formation but do not by themselves establish taxon-specific production rates. In sourdough, integrated metagenomic and metatranscriptomic analyses of eight spontaneous fermentations revealed both functional redundancy and species-specific specialization in carbohydrate, pyruvate, and nitrogen metabolism, with Lactiplantibacillus plantarum expressing maltose-metabolism genes, Furfurilactobacillus rossiae expressing beta-glucosidase and arabinoxylan-utilization functions, and Saccharomyces cerevisiae and Pichia kudriavzevii showing complementary transcriptional profiles for starch, dextrin, and fructose metabolism.

Cheese ripening research illustrates how metatranscriptomics resolves temporal dynamics that community DNA profiling cannot capture. In surface-ripened cheeses, transcripts associated with lactose utilization were assigned mainly to Lactococcus lactis and Kluyveromyces lactis, while later-stage lactate-utilization transcripts were associated with Debaryomyces hansenii and Geotrichum candidum, and amino-acid-catabolism transcripts were assigned to G. candidum, Corynebacterium casei, and Hafnia alvei. In Reblochon-style cheese, amino-acid-catabolism transcripts were upregulated in both dominant yeasts, earlier in G. candidum than in D. hansenii, while lactic acid bacterial transcriptomes showed fewer changes during ripening. In Chinese Baijiu production, comparisons among Daqu types indicate that temperature regime is associated with distinct RNA-derived taxonomic profiles, with fungal-associated transcripts predominating in low-temperature Daqu and stronger bacterial-associated signals in medium- and high-temperature varieties. During sauce-flavor Baijiu fermentation, early-phase transcripts assigned to fungi mapped predominantly to sulfur metabolism, glyoxylate and dicarboxylate metabolism, lipid metabolism, and the tricarboxylic acid cycle, while middle and later stages showed increasing bacterial representation in carbon, butanoate, propanoate, and amino acid metabolism pathways.

Across all systems, the review identifies recurrent transcriptional patterns alongside system-specific signatures. Lactic acid bacteria-associated carbohydrate catabolism and acidification recur in fermented vegetables, sourdough, and dairy products, while yeast-associated alcoholic fermentation and aroma pathways are more strongly represented in alcoholic beverages and also appear in sourdough and soybean fermentations. Proteolysis and amino acid catabolism recur in dairy, soybean, and alcoholic fermentations, where they may contribute either to desirable flavor development or to safety risks such as biogenic-amine accumulation. Sulfur- and lipid-metabolism transcripts concentrate in aroma-oriented systems such as Baijiu, fermented vegetables, and cheese, though their contributions to volatile formation often remain inferred rather than experimentally validated. Nitrite-metabolism transcripts have been reported mainly in fermented vegetables, while oxygen-dependent ethanol-oxidation pathways associated with acetic acid bacteria characterize vinegar and filamentous-fungal saccharification pathways distinguish Baijiu production.

The review’s central message is that metatranscriptomics should function as an early-warning and hypothesis-prioritization tool rather than as a standalone measure of metabolic activity or food-safety risk. Apparent conflicts among studies may reflect food-matrix and analytical effects as much as true biological divergence, since fermentation stage, RNA extraction protocols, sequencing depth, reference databases, and normalization strategies remain unstandardized. Progress depends on replicated time-series designs combined with matched metagenomic, metaproteomic, or metabolomic measurements, explicit reporting of rRNA depletion and retained microbial-mRNA fractions, biological replication, and transparent normalization. Machine learning may help prioritize poorly characterized proteins, but its predictions require experimental validation. Containerized workflows can improve computational reproducibility, and process-sensor data synchronized with transcript sampling could relate environmental variation to gene expression. These advances can inform the design of defined microbial consortia and candidate starter cultures, but predicted improvements must be demonstrated through targeted interventions before implementation in fermentation control, ensuring that transcriptional signals ultimately translate into measurable improvements in flavor, safety, batch consistency, and fermentation duration.

Subject of Research: Metatranscriptomic profiling of microbial gene expression in traditional fermented foods

Article Title: Harnessing metatranscriptomics to unveil the metabolic landscape of microbiota in traditional fermented foods

Article References: Li, R., Fan, X., Kang, J., Huang, X., Shi, Y., & Chen, X. (2026). Harnessing metatranscriptomics to unveil the metabolic landscape of microbiota in traditional fermented foods. Current Research in Food Science, Article 101601. https://doi.org/10.1016/j.crfs.2026.101601

Image Credits: AI Generated

DOI: 10.1016/j.crfs.2026.101601

Keywords: metatranscriptomics, fermented foods, microbiome, RNA sequencing, lactic acid bacteria, kimchi, cheese ripening, sourdough, Baijiu, biogenic amines, multi-omics, food safety