Traditional fermented food microbial ecosystems oriented towards industrial standards: multi-dimensional characterization and artificial regulation

Abstract

Fermented food microbial ecosystems are dynamic and shaped by microbes, substrates, and environments, but traditional spontaneous fermentation causes variability in quality and safety, limiting standardization. This review highlighted multi-omics integration to comprehensively characterize microbial composition, dynamics, interactions, metabolites, and phages, and summarized regulation strategies including functional strains, synthetic microbial communities, bacteriophage control, and physicochemical modulation. Finally, the futures and challenges from empirical practice to predictable, scalable industrial fermentation were proposed.

Introduction

Fermented foods are defined as those made through desired microbial growth and enzymatic conversions of food components¹. Fermented foods have a long and rich history, with origins dating back to as early as 7000 BCE. They have remained a crucial part of the human diet, accounting for approximately one-third of daily food intake^1,2. In recent years, the demand for fermented foods increasing at a rate of one to three times³. Based on the raw materials, fermented foods can be categorized into several main types, including fermented cereal products, dairy products, meat products, vegetable products, and alcoholic beverages. Compared to non-fermented foods, fermented foods are characterized by distinctive flavors, textures, appearances, and functional properties, which significantly enhance their taste, nutritional value, and economic worth⁴. Currently, many fermented products, such as beer and yogurt, are produced on an industrial scale using well-defined, single or simple starter cultures⁵. This approach ensures a controlled fermentation process and stable product quality. However, while industrial fermentation ensures consistency, the unique flavor profiles of many traditional fermented foods often arise from their open, complex microbial ecosystems. Nevertheless, many traditional fermented foods, such as vinegar, sauerkraut, kefir, and Baijiu (Chinese liquor), continue to rely on traditional fermentation processes that occur spontaneously in open environments⁶. Furthermore, with the advent of Industry 4.0 and the improvement of living standards, the consumer preference and social requirement have shifted toward standardized production of traditional fermented foods³. As a result, the modernization of traditional fermentation processes is an urgent priority, balancing the need for consistency with the preservation of the distinctive sensory characteristics that define these foods.

The microbial ecosystem plays a crucial role in the production of fermented foods, directly influencing the flavor quality, safety, and stability of the final product^7,8. Therefore, scientifically analyzing and precisely controlling the fermentation microbial ecosystem will facilitate the standardization of traditional fermented food production. The microbial ecosystem of traditional fermented foods is a dynamic fermentation system formed through the interaction of microbial communities (such as molds, yeasts, bacteria, actinomycetes, etc.), the raw materials and the surrounding environment⁹. The microbial community includes a diverse array of microorganisms, which can originate from the raw materials (e.g., wine, and ham)^10,11, complex starter cultures (e.g., Baijiu, and vinegar)^12,13, or from previous fermentation batches (e.g., back-slopping in sourdough, sauerkraut, and kefir)^14,15. The collaborative division of labor and resource exchange between these complex microorganisms contributes to the resilience, adaptability, and stability of the fermentation microbial ecosystem¹⁶. Furthermore, the microbial ecosystem fluctuates over time and space during fermentation due to factors such as raw material composition, fermentation methods, environmental parameters, disturbances from exogenous microorganisms, and interactions among the microorganisms themselves^7,17. Although the central role of the microbial ecosystem is well-recognized, its complexity, dynamism, and responsiveness to various environmental factors present significant challenges for systematic analysis and precise regulation¹⁸.

Studies about the microbial ecosystem of fermented foods has traditionally been limited to cultivation-based techniques, which focus on isolating and characterizing the functions of individual microorganisms¹⁹. These methods lack the ability to analyze microbial community assembly, functional metabolism, and their relationship with product quality within the entire microbial ecosystem. However, with the rapid advancements in molecular biology and high-throughput sequencing technologies (including amplicon sequencing, genomics, transcriptomics, proteomics, and metabolomics), sequencing costs have significantly decreased, and public databases have expanded^20–23. As a result, scientists have gained a broader understanding of the microbial ecosystems in traditional fermented foods, including microbial species, diversity, gene expression, gene function, and metabolites^24,25. However, despite the remarkable progress enabled by multi-omics technologies in elucidating the microbial ecosystems of traditional fermented foods, substantial challenges remain in translating multi-dimensional omics datasets into actionable and controllable process regulation strategies. Integrative multi-omics analysis requires the effective fusion of large-scale information across multiple biological layers, including microbial community structure, functional gene expression, metabolic pathway activity, and dynamic metabolite profiles. These datasets must then be interrogated using systems biology models, machine learning algorithms, and other computational approaches to identify key functional microorganisms, critical interaction networks, and potential regulatory control nodes. Nevertheless, the practical implementation of multi-omics-driven fermentation control still constrained by several major limitations. These include limited reproducibility due to the inherent complexity and variability of fermentation microbial ecosystems, as well as difficulties in integrating heterogeneous data from different omics platforms. Moreover, there is a lack of direct and standardized mappings between omics-derived indicators and the critical process parameters needed for robust industrial fermentation regulation²⁶. In this context, industrial standards refer to the systematic frameworks enabling mass production with batch-to-batch consistency, process controllability, and predictable product quality, but specifically emphasizing the technological capability for precise microbial ecosystem regulation that ensures sensory heritage preservation alongside safety and uniformity.

In recent years, growing attention has been directed toward the microbial ecosystems of fermented foods. Nevertheless, most existing reviews have largely remained at the level of descriptive characterization, primarily emphasizing microbial community composition and diversity patterns. They provide insufficient discussion on multi-omics-enabled multidimensional profiling of fermentation ecosystems. More importantly, these reviews also lack adequate exploration of how such insights can be translated into practical, industry-oriented strategies for process regulation and standardization. Notably, bacteriophages, as pivotal ecological drivers shaping microbial community stability and succession, have been largely neglected in previous reviews, and their functional roles and regulatory potential within fermentation microbial ecosystems have yet to be comprehensively synthesized and critically evaluated. Based on it, this review will provide an overview of recent advancements in the multi-dimensional characterization of fermented food microbial ecosystems using multi-omics technologies, including deciphering microbial composition and community structure, spatiotemporal succession dynamics, microbial interactions, metabolic profiles, and bacteriophage populations using multi-omics strategies. Furthermore, it also summarizes methods and strategies for artificially regulating these microbial ecosystems in recent years and offer insights into achieving a deeper, systematic understanding of traditional fermentation microbial ecosystems for precise regulation. Integrating these insights will contribute to a systematic framework for precise fermentation control, thereby facilitating standardized quality enhancement and accelerated industrialization. And in this context, industrial standards refer to the systematic frameworks enabling mass production with batch-to-batch consistency, process controllability, and predictable product quality, but specifically emphasizing the technological capability for precise microbial ecosystem regulation that ensures sensory heritage preservation alongside safety and uniformity.

Multi-dimensional characterization of microbial ecosystems with the development of emerging technologies

With the rapid advancement of high-throughput sequencing, multi-omics technologies, and bioinformatics, the multi-dimensional characterization of microbial ecosystems in traditional fermented foods has emerged as a major research focus. Recent studies have primarily concentrated on elucidating microbial community composition and structure during fermentation, quantitatively profiling microbial populations, investigating the dynamic changes of microbial metabolites, exploring inter-microbial interactions, and the bacteriophages in microbial community. These efforts contribute to a deeper understanding of the intrinsic mechanisms underlying traditional fermented foods and provide a solid theoretical foundation for fermentation process regulation and industrial upgrading (Fig. 1). Specifically, high-throughput sequencing technologies are primarily employed to elucidate the microbial community composition and structure during fermentation, characterize spatiotemporal succession patterns, and identify core microbiota based on abundance dynamics and correlation-based network analyses, thereby providing a fundamental basis for constructing dynamic microbial community models. Transcriptomic profiling further captures the expression levels of key genes and regulatory networks within microbial communities, enabling the elucidation of metabolic activity shifts across different fermentation stages. Metabolomics, in turn, directly reflects the production and accumulation of microbial metabolites, including flavor compounds, bioactive substances, and potentially hazardous metabolites, thus establishing a direct link between microbial functional dynamics and product quality attributes. Meanwhile, absolute quantification approaches can correct biases inherent to relative abundance measurements, generating data that more closely approximate true microbial loads and substantially improving the accuracy and comparability of downstream models. Building upon these datasets, integrative multi-omics analysis—via correlation analysis, causal inference, dynamic network reconstruction, or machine learning-based modeling—can establish a systematic association framework spanning “community structure–functional expression–metabolic outputs–quality phenotypes”. This integrative pipeline facilitates the identification of key functional microorganisms, critical interaction modules, and pivotal metabolic nodes that govern fermentation stability and quality formation. Ultimately, such a framework provides an operational foundation for fermentation process regulation, supporting real-time monitoring, risk early warning, and dynamic intervention strategies, thereby promoting the transition of traditional fermentation from experience-driven practices toward data-driven and standardized industrial production.

Fig. 1. Systematic characterization of microbial ecosystems through emerging technologies.

a Insights into microorganism composition and structure. b Revealing microbial interactions. c Resolving microbial metabolites. d Deciphering bacteriophage and phage-bacteria interaction.

Microbial composition and structure

Main kinds of microorganisms and their functions

The primary microorganisms involved in the production of traditional fermented foods include molds, yeasts, lactic acid bacteria (LAB), acetic acid bacteria, and Bacillus species (Table 1). Molds produce key enzymes such as amylases, proteases, and lipases. Amylases break down complex polysaccharides in fermented substrates into smaller molecules like glucose, providing essential carbon sources and energy for other microorganisms. Proteases hydrolyze proteins into amino acids, and enhances the umami flavor of the product²⁷. Additionally, lipases catalyze esterification reactions, producing aromatic esters that significantly improve the sensory attributes of fermented food²⁸. As a model microorganism, Saccharomyces cerevisiae is a dominant microbial member in traditional fermented food ecosystems. It excels in rapidly converting sugars into ethanol under anaerobic, high-sugar, and low-pH conditions, engaging in intense interactions with other microbial strains during the fermentation process²⁹. In contrast, non-Saccharomyces yeasts are generally less ethanologenic but critically contribute to the formation of distinct flavor compounds, including phenylethanol (rose-like), phenylethyl acetate (honey-like), and furfural (caramel-like aroma)³⁰. These compounds play an integral role in shaping the unique flavor profiles of fermented foods. Lactic acid bacteria are pivotal in many traditional fermented foods. They metabolize carbohydrates into lactic acid, lowering the pH and thereby suppressing the proliferation of harmful microorganisms. The lactic acid itself, along with flavor compounds such as ethyl lactate, generated as byproducts, significantly contributes to the organoleptic qualities of the product. Moreover, metabolites from LAB often enhance the nutritional profile of fermented foods, thereby transforming them from mere flavoring agents into functional foods with tangible health benefits^31,32. Acetic acid bacteria, which function in aerobic fermentation environments, produce acetic acid and are indispensable in the products like vinegar, kefir, and sourdough. These bacteria are also instrumental in enhancing both the flavor and quality of the fermented foods^33–35. Bacillus subtilis, a key species within its genus, serves distinct roles in traditional Asian fermented foods. In Chinese soy sauce-flavored Baijiu, it contributes crucially as a producer of characteristic flavor compounds like pyrazines. Conversely, in Japanese natto fermentation, it plays a central role by generating not only flavor compounds but also functionally important bioactive substances³⁶.

Table 1.

The composition of main microorganisms of classical fermented foods

	Fermentation food	Raw materials	State of fermentation	Main microorganisms	References
Fermented grain products	Soy Sauce	Soybean	Semi-solid	Mold: Aspergillus Yeast: Candida; Zygosaccharomyces rouxii Bacteria: Weisella, Bacillus, Lactobacillus	36
	Bean Paste	Soybean	Solid	Mold: Aspergillus oryzae Yeast: Zygosaccharomyces Bacteria: Tetragenococcus, Lactobacillus, Staphylococcus, Acinetobacter, Pseudomonas, Streptococcus	150
	Vinegar	Wheat, sorghum, corn, wheat bran, rice	Solid	Mold: Eurotium, Monascus, Aspergillus Yeast: Pichia, Saccharomyces, Saccharomycopsis Bacteria: Acetobacter, Saccharopolyspora, Bacillus, Weissella, Lactobacillus, Lactococcus, Gluconacetobacer	151
	Sour dough	Wheat flour	Solid	Yeast: Saccharomyces cerevisiae Bacteria: Lactobacillus, Acinetobacter, Pantoea	40
	Sufu	Soybean	Solid	Mold: Actinomucor, Mucor, Rhizopus Bacteria: Micrococcus, Bacillus	152
Alcoholic beverages	Baijiu	Sorghum	Solid	Bacteria: Acetobacter, Bacillus, Enterococcus, Lactobacillus, Leuconostoc, Pediococcus, Weissella. Yeast: Candida, Pichia, Saccharomyces, Wickerhamomyces. Mold: Aspergillus, Geotrichum, Rhizopus.	153
	Grape wine	Grape	Liquid	Bacteria: Oenococcus oeni Yeast: Hanseniaspora, Metschnikowia, Pichia, Torulaspora	154
	Rice wine	Rice	Semi-solid	Bacteria: Bacillus, Enterobacter, Lactobacillus, Lactococcus, Leuconostoc, Pseudomonas, Saccharopolyspora, Staphylococcus, Thermoactinomyces, Weissella Yeast: Candida, Cryptococcu, Epicoccum, Saccharomyces Mold: Aspergillus, Fusarium, Rhizomucor, Thermomyces	155
Fermentated dairy products	Kefir	Milk	Liquid	Mold: Aspergillus amstelodami Bacteria: Lactobacillus kefiranofaciens, Lactococcus lactis, Leuconostoc mesenteroides, Lactobacillus kefiri, Lactobacillus helveticus, Acetobacter okinawensis, Acetobacter orientalis, Enterobacter aerogene. Yeast: Kazachstania unispora, Kazachstania turicensis, Kluyveromyces marxianus, Saccharomyces cerevisiae	156
	Koumiss	Mare milk	Liquid	Bacteria: Lactobacillus helveticus, Lactobacillus kefiranofaciens, Lactobacillus plantarum, Lactobacillus buchneri, Lactobacillus salivarius, Lactobacillus rhamnosus; Leuconostoc mesenteroides subsp. cremoris Yeast: Kluyveromyces marxiamus, Pichia fermentans, Saccharomyces cerevisiae	157
	Cheese	Milk	Liquid and solid	Mold: Geotrichum candidum, Penicillium camemberti, Penicillium roqueforti Bacteria: Lactococcus lactis, Lactobacillus delbruckii, Lactobacillus helveticus, Lactobacillus plantarum, Streptococcus thermophilus, Enterococcus durans, Enterococcus faecium, Brevibacterium linens, Propionibacterium freudenreichii Yeast: Debaryomyces hansenii	158
Fermented fish and meat products	Ham	Pork	Solid	Mold: Penicillium italicum, Aspergillus sydowi, Aspergillus flauipes Yeast: Hansenula sydowiorum, Debaryomyces hansenii, Hansenula polymorpha Bacteria: Pediococcus pentosaceus, Streptococcus equi, Staphylococcus equorum, Staphylococcu sequourm, Pediococcus urinacequi	159
	Fermented sausage	Pork	Solid	Bacteria: Lactobacillus, Weissella, Leuconostoc, Staphylococcus Yeast: Debaryomyces hansenii	160
	Surstrmming	Herring	Solid	Bacteria: Halanaerobium praevalens, Alkalibacterium gilvum, Carnobacterium spp., Tetragenococcus halophilus, Clostridiisalibacter spp., Porphyromonadaceae, Psychrobacter celer, Marinilactibacillus psychrotolerans, Streptococcus infantis, Salinivibrio costicola, Ruminococcaceae	161
	stiky Mandarinfish	Mandarinfish	Solid	Bacteria: Lactobacillus sakei, Streptococcus parauberis, Lactococcus garvieae, Lactococcus lactis, Enterococcus hermanniensis, Macrococcus caseolyticus, Coldbacterium, Enterococcus	162
Fermented vegetable products	Fuling zhacai	Mustard tuber, salt	Semi-dry	Bacteria: Companilactobacillus alimentarius, Halomonas spp., Lactobacillus brevis, Lactobacillus curvatus, Ldiomarina loihiensis Yeast: Debaryomyces hansenii	163
	Paocai	Cabbage, salt	Semi-solid	Bacteria: Enterococcus, Gluconacetobacter, Lactococcus, Leuconostoc, Pectobacterium	164
	Suancai	Cabbage, salt	Semi-solid	Bacteria: Enterococcus, Lactobacillus, Pediococcus	165
	Kimchi	Cabbage, radish, salt, spices	Semi-solid	Bacteria: Lactobacillus, Leuconostoc, Pediococcus, Weissella	166,167

The aforementioned major microbial groups formed the basic framework of fermentation, but it was the advent of high-throughput sequencing technologies that truly revealed the astonishing diversity and dynamic nature of microbial ecosystems. Currently, most studies employ 16S rRNA and ITS (Internal transcribed spacer) amplicon sequencing to elucidate the microbial community composition and succession during the production of traditional fermented foods. For instance, Xu et al. employed 16S rRNA and ITS amplicon sequencing to profile the microbial communities during the first three fermentation cycles of soy sauce-flavor Baijiu. Their analysis revealed that the bacterial communities in the first, second, and third cycles consisted of 229, 235, and 205 genera, respectively, and shared a core microbiome of 160 genera. Similarly, fungal diversity comprised 141, 169, and 209 genera across the cycles, with 106 genera common to all³⁷. Despite the impressive microbial diversity revealed by high-throughput sequencing, not all microorganisms contribute functionally to fermentation outcomes. Studies have shown that the composition and function of the core microorganisms in a complex fermentation microbial ecosystem largely determine the fermentation direction and product quality^22,38. Core microorganisms are defined by their crucial functional roles in community assembly and dynamic succession during fermentation, rather than by their relative abundance³⁹. Consequently, the identification of core functional microbiota has become a research focus. A case in point is the work by Wang et al., who employed an O2PLS (bidirectional orthogonal partial least squares) correlation model, integrating 16S rRNA sequencing data with microbial succession and flavor dynamics during Zhenjiang vinegar fermentation. This approach pinpointed seven key genera, Acetobacter, Lactobacillus, Enhydrobacter, Lactococcus, Gluconacetobacter, Bacillus, and Staphylococcus, as crucial drivers of the vinegar fermentation⁴⁰. Unlike methods that only profile taxonomic composition, metatranscriptomics identifies genes being actively expressed within a community at specific times, providing a functional blueprint of metabolic activity. For example, by applying this approach to soy sauce-flavor Baijiu fermentation, Song et al. pinpointed a core functional microbiota comprising 10 fungal and 11 bacterial genera, establishing a direct link between microbial transcription and process dynamics³⁸. Moreover, microbial succession during fermentation has been frequently elucidated through temporal sampling. In a study on soy sauce fermentation, Zhang et al. employed high-throughput sequencing to analyze microbial community dynamics. Weissella, Tetragenococcus, and Lactobacillus were identified as the marker bacteria for the initial, middle, and late stages, respectively. While Aspergillus predominated throughout, Candida and Zygosaccharomyces emerged predominantly later. Critically, correlation analyses integrating microbial data with flavor compounds and physicochemical profiles identified Tetragenococcus, Lactobacillus, and Zygosaccharomyces as potential core functional genera, linking them to environmental modulation and characteristic flavors formation⁴¹.

Methods for investigating microbial diversity from relative quantification to absolute quantification

The production of traditional fermented foods occurs under open conditions, making the fermentation process susceptible to various unpredictable factors⁴². For instance, human intervention and environmental imbalances can lead to differential distribution of microbial communities across distinct spatial niches. Therefore, in addition to investigating the temporal dynamics of microbial community shifts, exploring spatial variations in community structure is pivotal for understanding microbial assembly and metabolic regulation mechanisms. Qian et al. employed real-time quantitative PCR to assess the biomass in jiupei and pit mud samples during the fermentation of strong-flavor Baijiu⁴³. Subsequent cluster analysis and principal component analysis revealed significant differentiation between the bacterial communities in the jiupei and pit mud. Functional predictions through PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) further indicated that enzymes involved in the biosynthesis of acetic acid and lactic acid were primarily concentrated in the jiupei samples, while the pit mud microbiome displayed greater potential for butyric acid and hexanoic acid synthesis. In vitro simulated fermentation investigation corroborated these findings, demonstrating that the jiupei microbial community is responsible for generating acetic acid and lactic acid, which are subsequently metabolized by the pit mud microbiome into butyric acid and hexanoic acid. This interplay between the jiupei and pit mud microbial communities contributes significantly to the formation of the characteristic flavor profile in strong-flavor Baijiu⁴³. Zhao et al. investigated the microbial community dynamics during the fermentation of Doubanjiang (a fermented broad bean paste) and identified Staphylococcus and Bacillus as the predominant bacterial genera, with Aspergillus and Zygosaccharomyces dominating the fungal population. Due to the low salinity tolerance, Aspergillus activity was largely suppressed in the later fermentation stages. In contrast, the other three dominant genera exhibited distinct spatial stratification, such as Staphylococcus and Zygosaccharomyces were more abundant in the upper fermenting mash, while Bacillus was confined to the lower layers. This niche differentiation was driven by environmental gradients (e.g., light, moisture) and microbial interactions. Systematic spatiotemporal analysis linked this spatial heterogeneity to final product quality, showing that it facilitated the recycling of degradation products and metabolic intermediates, thereby maintaining community homeostasis⁴⁴. This scientific exploration of the empirical practices within traditional fermentation processes offers valuable insights, providing a robust theoretical framework to guide the modernization of fermented food production.

Despite amplicon sequencing widespread use in fermentation microbiome studies, a key limitation is that it merely depicts relative microbial abundances and cannot discern absolute quantities, thus providing an incomplete and potentially misleading view of community structure. This fundamental constraint may lead to the oversight or misjudgment of microorganisms critical to the fermentation process. As a result, absolute quantification approaches have attracted growing attention for their abilities to provide a more comprehensive understanding of microbial community structures^45,46. To achieve absolute quantification, Du et al. leveraged two abundant and ubiquitous species in Baijiu fermentation, Lactobacillus acetotolerans and Lactobacillus jinshani, as endogenous internal references. The absolute abundance of these reference organisms was first quantified via species-specific qPCR. Subsequently, amplicon sequencing data were normalized on the basis of these qPCR measurements to estimate the absolute abundance of the entire bacterial community⁴⁷. In addition, Wang et al. introduced the Gradient Internal Standard Absolute Quantification (GIS-AQ) method. This approach entails spiking a single reference sample with a concentration gradient of internal standards, enabling precise quantification of microbial taxa across several orders of magnitude. The accuracy of the GIS-AQ method was further validated through microscopic quantification⁴⁸. Wang et al. applied this method to analyze the absolute abundance of microorganisms during both the first and second fermentation rounds of jiupei, while also examining the relative abundance of microbial populations at different stages of baijiu fermentation. Their findings revealed marked discrepancies in the microbial abundance profiles obtained from the two methods. Importantly, the absolute abundance data provided insights into key functional microbial communities involved in starch degradation, thus offering valuable targets for improving the efficiency of fermentation substrate utilization⁴⁹. Hu et al. developed a one-step Recombinase Polymerase Amplification (RPA)-CRISPR/Cas12a (ORCC) method for the rapid and quantitative detection of three critical yeasts, Saccharomyces pastorianus, Saccharomyces cerevisiae, and Zygosaccharomyces bailii, commonly used in fermented foods. This method integrates RPA for rapid isothermal DNA amplification at 37 °C with CRISPR/Cas12a-mediated detection. In this system, Cas12a, guided by a specific crRNA, cleaves the target DNA and simultaneously activates non-specific trans-cleavage activity, which degrades fluorescently labeled single-stranded DNA probes, resulting in the release of a fluorescent signal. Quantitative detection is achieved by correlating the fluorescent signal with the concentration of the target DNA. This streamlined approach reduces the overall detection time to just 45 min. Validation experiments demonstrated that the ORCC method offers superior sensitivity, specificity, and a lower detection limit in comparison to qPCR. Moreover, it can successfully detect target yeasts in both pure cultures and natural fermentation samples⁵⁰.

Microbial metabolites

Microbial activity ultimately determines the quality of fermentation products through its metabolic byproducts. Therefore, systematically characterizing metabolites is crucial for bridging microbial activity and the resulting properties of fermentation products. During fermentation, microbial growth and metabolism yield enzymes, water, and a diverse array of metabolites comprising both volatile and non-volatile compounds. Elucidating the formation and feedback mechanisms of metabolites within fermentation-associated microbial ecosystems is essential for rationally optimizing fermentation processes. In a study by Wen et al., untargeted metabolomics was employed to characterize metabolic profiles across different stages of soy sauce fermentation. This approach led to the identification of 60 non-volatile metabolites, including 27 peptides, amino acids, and derivatives; 8 carbohydrates and related complexes; 14 organic acids and derivatives; 5 amides; 3 flavonoids; and 3 nucleosides. Furthermore, through integrated metagenomic and bioinformatic analyses, the formation and degradation pathways of 13 free amino acids were delineated, providing valuable insights for the targeted regulation of soy sauce production⁵¹. Similarly, in a study of Korean kimchi, Zhao et al. employed metabolomics to identify eight key flavor-active and sixteen aroma-active compounds central to its distinctive sensory profile. By integrating metagenomic analysis, the researchers reconstructed a metabolic network that delineates the biosynthetic pathways of these signature compounds, thereby establishing a theoretical basis for the targeted optimization of fermentation processes⁵².

Microbial metabolites not only serve as byproducts of fermentation but also exert feedback effects on the microbial ecosystem. During fermentation, physicochemical conditions undergo continuous alterations in tandem with the accumulation of microbial metabolites, including pH, ethanol, reducing sugars, dissolved oxygen, and other key parameters. Such dynamic changes impose intense ecological filtering pressure within the fermentation microbial ecosystem, thereby selectively enriching microbial taxa endowed with adaptive metabolic traits⁵³. Concurrently, progressive substrate depletion and niche differentiation further drive microbial competition and functional replacement, resulting in distinct successional shifts in community structure across fermentation stages. Taking Shanxi aged vinegar as a model system, Wu et al. applied meta-transcriptomics to resolve the dynamic metabolic pathways of organic acids throughout fermentation. Their analysis pinpointed active microorganisms and their specifically expressed enzymes, revealing the regulatory role of core microorganisms in sustaining continuous fermentation. In the early and mid-fermentation phases, Lactobacillus predominantly drove the production of acetic and lactic acids, whereas Acetobacter emerged as the dominant producer in later stages. Notably, genes such as acs and aarC in Acetobacter were implicated in acetic acid assimilation. As acidity pressure increased, the fermentation microbial ecosystem shifted from Lactobacillus to Acetobacter dominance, thereby ensuring the progression and stability of vinegar fermentation⁵⁴. In the study of Hao et al., they linked microbial dynamics with fermentation parameters in sauce-flavor baijiu to elucidate metabolite-mediated ecosystem regulation. Their results identified reducing sugars as the primary driver of microbial succession during stacking fermentation, where high sugar levels promoted growth and positively correlated with dominant genera such as Virgibacillus, Kroppenstedtia, and Pediococcus. In contrast, cellar fermentation was primarily governed by microbial-derived ethanol and organic acids. The enclosed pit environment also led to heat accumulation, making temperature a critical modulating factor. During this stage, ethanol, acidity, and temperature showed positive correlations with the dominant microorganisms Lactobacillus and Saccharomyces⁵⁵. Furthermore, Zhao et al. integrated metagenomic and metaproteomic analyses to profile the microbial communities involved in the solid-state fermentation of Pu-erh tea. Their study identified key fermentation-associated enzymes and established Aspergillus as the dominant fungal genus and primary protein contributor. These insights significantly deepen the mechanistic understanding of flavor compound formation during tea fermentation⁵⁶. Moreover, the volatile compounds produced during fermentation also influence microbial metabolism within the microbial ecosystem. For example, Zhang et al. integrated high-throughput sequencing, metabolomics, transcriptomics, and gas chromatography-mass spectrometry to analyze the impact of volatile compounds on baijiu fermentation. Their results revealed that volatile compounds, particularly 2-phenylethanol produced by Pichia spp., inhibited the germination and hyphal growth of filamentous fungi both directly and indirectly. This antifungal activity primarily occurs through the disruption of fungal protein synthesis and DNA damage. The inhibition of filamentous fungi helps maintain yeast diversity, thereby promoting ethanol metabolism and improving the flavor and fermentation efficiency of baijiu⁵⁷. These studies underscore the pivotal role of metabolites, they not only as products of fermentation but also as critical feedback signals that regulate the dynamic changes of microbial community within the microbial ecosystem.

Nevertheless, despite the growing abundance of spatiotemporal sequencing data, inferring the underlying ecological mechanisms solely from microbial successional patterns and changes in physicochemical properties remains considerably challenging. First, relative abundance profiling based on amplicon sequencing is susceptible to multiple confounding factors, including compositional effects, as well as biases in DNA extraction and PCR amplification introduced by the complex matrix of fermentation samples. Such interferences may lead to erroneous interpretations of microbial community succession during fermentation. In addition, the timing of sampling throughout the fermentation process is critical, as key transitional stages often govern the shift in fermentation direction. This necessitates multi-point sampling across both temporal and spatial scales when investigating microbial community assembly in fermentative microbial ecosystems, thereby facilitating a comprehensive elucidation of the embedded ecological mechanisms. Accordingly, future research demands high-resolution spatiotemporal dynamic sampling and the establishment of standardized experimental protocols to uncover more robust and reproducible ecological principles governing fermentation processes.

Microbial interactions

After elucidating “who is present” (community composition) and “what they do” (metabolites), understanding “how microorganisms interact” (inter-species microbial relationships) also vital to decipher the assembly rules and functional mechanisms of microbial ecosystems. Changes in keystone species within microbial communities can lead to pronounced shifts in community structure, including fluctuations in microbial diversity and relative abundance. This indicated that keystone microorganisms engage in extensive interactions both with one another and with other community members within fermented microbial ecosystems. Microbial interactions can be broadly classified into mutualism (+ +), commensalism (+0), antagonism (+−), neutralism (0 0), amensalism (0 −), and competition (−−)⁵⁸. Mutualistic relationships are especially prevalent within the complex microbial consortia responsible for food fermentation. Current studies on microbial interactions have predominantly centered on pairwise models between two species, through which three principal interaction mechanisms have been delineated as follows.

Quorum sensing (QS) is an intercellular signaling mechanism that depends on cell density. As cell density rises, QS signal molecules accumulate to a threshold level, where they can be detected by intracellular target receptors. This recognition subsequently initiates coordinated gene expression and physiological activities^59,60. Beyond its regulatory effects on microorganisms, such as influencing cell behavior, stress tolerance, and the production of secondary metabolites, quorum sensing also plays a crucial role in inter-microbial interactions⁶¹. For instance, during co-cultivation with lactic acid bacteria isolated from Koumiss (a fermented mare’s milk), Saccharomyces cerevisiae YE4 suppressed the expression of quorum-sensing genes (luxS and pfs in Enterococcus faecium 8-3 and luxS in Lactobacillus fermentum 2-1). Conversely, YE4 concurrently enhanced its own secretion of autoinducer-2 (AI-2) and upregulated the expression of its associated quorum-sensing molecular genes^62,63. Additionally, Zeng et al. demonstrated that the signaling molecule comQXPA produced by Bacillus velezensis isolated from fermented grain samples could inhibit the growth of Saccharomycopsis fibuligera, thereby influencing the production of ester compounds in baijiu fermentation⁶⁴. Despite these insights, a significant gap remains in understanding quorum sensing among microorganisms in fermentation ecosystems, with this complex interaction network still largely unexplored and in its early stages of investigation.

The exchange of microbial metabolites and nutrients leads to changes in gene expression in co-cultured microorganisms during fermentation^65,66. In ethanol-fermented dairy products like kefir and koumiss, the core microbiota, Saccharomyces cerevisiae and lactic acid bacteria (LAB), constitute a classic model of microbial interaction. Since Saccharomyces cerevisiae cannot metabolize lactose, LAB hydrolyze this primary carbon source into galactose and glucose, thereby supplying carbon to the yeast. Conversely, transcriptomic analyses reveal a reciprocal regulatory network, that is lactic acid from LAB chelates metal cations, inducing overexpression of iron/copper transporter genes in yeast, while yeast-derived ethanol stimulates LAB genes involved in long-chain fatty acid metabolism. This demonstrates how cross-feeding of metabolites can drive species-specific gene expression and reshape metabolic pathways^67,68. Blasche et al. emphasized that the structure of the kefir microbiome is governed by a dynamic equilibrium between mutualism and competition. Although Lactobacillus kefiranofaciens often dominates this community, it cannot thrive in milk monoculture due to metabolic dependence on Lactobacillus mesenteroides. Their cooperation is bidirectional: proteolytic activity by Lactobacillus kefiranofaciens supplies amino acids and peptides to Lactobacillus mesenteroides, while lactic acid produced by the latter stimulates growth of the former. Moreover, Lactobacillus kefiranofaciens synthesizes the exopolysaccharide matrix of kefir grains, which reinforces its competitive advantage and helps maintain its dominance throughout fermentation¹⁶.

The formation of microbial biofilms alters the survival patterns of microorganisms^69,70. In a microbial fermentation ecosystem, cells secrete extracellular polymeric substances, which aggregate to form co-aggregated biofilms. The biofilm act as a protective barrier, isolating the cells from harmful environmental factors and mitigating the stress imposed by external pressures⁷¹. Hirayama et al. characterized the stratified biofilm architecture formed by co-cultures of Lactobacillus plantarum ML 11-11 and Saccharomyces cerevisiae, both isolated from Fukuyama pot vinegar, using fluorescence in situ hybridization (FISH) and atomic force microscopy. The biofilm exhibited a two-layer organization, with Lactobacillus plantarum localized predominantly in the basal layer and both species coexisting in the upper layer, conferring increased structural robustness. Furthermore, the biofilm conferred protection against environmental stress to the lactic acid bacteria, thereby modulating their survival and metabolic profiles⁷². Kang et al. characterized the composition, structure, and microbial interactions of biofilms on Chinese baijiu fermentation containers. The core biofilm community was comprised of lactic acid bacteria, yeasts, and filamentous fungi. Upon isolating robust biofilm-forming and fermentative strains of lactic acid bacteria and yeast, co-culture experiments demonstrated a propensity for dual-species biofilm formation. This mixed-species biofilm exhibited marked differences in metabolic activity compared to their mono-species counterparts⁷³. And Zeng et al. indicated that the formation of biofilm in Saccharomycopsis fibuligera could change the synthesis of esters in Baijiu, more kinds of ester substances were synthesized, but less total ester content. They speculated that hydrolysates glucose and amino acids of EPS (Extracellular polymeric substances) may be used as precursors for the synthesis of esters and other flavor substances, but it needs to be further revealed⁶⁴.

However, higher-order interactions are far more prevalent than pairwise microbial interactions alone in fermentation microbial ecosystems. Such higher-order interactions typically manifest as intricate ecological networks, which confer stability to the fermentation microbial ecosystem and render species coexistence robust against perturbations in population abundance and environmental parameters⁷⁴. Nevertheless, current efforts to decipher these higher-order interaction mechanisms rely predominantly on correlation networks constructed from shifts in microbial relative abundance. Such computational approaches are highly sensitive to technical and data-processing factors, including sequencing depth and data normalization, and their outputs may not faithfully represent genuine biological interactions. Furthermore, correlation-based analyses fail to uncover causal relationships or the directionality of microbial interactions, while potentially overlooking indirect cross-talk among community members. Additionally, the in silico inferences of higher-order interactions remain largely unsupported by robust experimental validation. Accordingly, future research should integrate causal inference methodologies with systematic experimental validation frameworks, thereby elevating omics-derived interaction networks to a level that enables mechanistic interpretation and facilitates engineering applications in fermentation microbial ecosystems.

Bacteriophage and bacteriophage–bacteria interaction

Beyond bacteria-fungi interactions, bacteriophages act as microbial predators and constitute indispensable “regulators” within fermentation microbial ecosystems. Bacteriophages are viruses that specifically infect bacteria and exhibit strict host specificity, they are incapable of growth or replication once separated from their host cells. Bacteriophages in fermented microbial ecosystems can reshape community structure through predator–prey dynamics. However, their impacts are context-dependent: phage predation may enhance the stability and resilience of the microbial ecosystem, yet it can also disrupt fermentation processes and ultimately lead to fermentation failure⁷⁵.

In fermentation microbial ecosystems, deleterious phages can markedly reduce host bacterial populations, thereby freeing ecological niches and nutrients, and ultimately reshaping microbial community structure and diversity. However, excessive phage infection may severely disrupt fermentation processes, leading to delayed or arrested fermentation, compromised product quality and substantial economic losses^76,77. Meanwhile, the open environment of traditional food fermentation production further increases their vulnerability to phage intrusion⁷⁸. Ghosh et al. reported widespread contamination by Bacillus subtilis phages in Korean soybean-based fermented foods and the raw materials, with phage titers reaching as high as 3.7 × 10⁴ PFU g⁻¹. From these contaminated samples, 15 phages were isolated, among which BSP18 displayed high specificity toward Bacillus subtilis and a relatively broad host infection spectrum. Notably, even artificial inoculation with a low concentration of BSP18 (1 × 10² PFU g⁻¹) was sufficient to significantly suppress Bacillus subtilis growth during soybean fermentation, and γ-polyglutamic acid was readily degraded by hydrolases encoded by BSP18, resulting in slowed fermentation dynamics and a decline in product quality⁷⁹. In addition, phage populations within fermentation microbial ecosystems are not static but undergo continuous evolution. Aaron et al. analyzed metagenomic data from 328 cheese samples and identified clear evidence of phage-mediated horizontal gene transfer. The widespread occurrence of CRISPR-based anti-phage defense systems in these metagenomes indicates that cheese-associated bacteria have evolved effective strategies to counter phage infection. In response, phages have co-evolved anti-CRISPR proteins to circumvent these bacterial defenses⁸⁰. Similarly, Dupont et al. isolated 936-type phages from whey samples that were capable of cross-infecting Lactococcus lactis ssp. lactis, Lactococcus lactis ssp. cremoris, and Lactococcus lactis ssp. diacetylactis, demonstrating an expanded and more dynamic host range. This finding contrasts with earlier reports, indicating that most 936-type phages exhibited restricted infectivity toward either Lactococcus lactis ssp. cremoris or Lactococcus lactis ssp. lactis^81,82. Collectively, these observations underscore the ongoing evolutionary arms race between bacteria and bacteriophages within fermentation microbial ecosystems.

A detailed understanding of bacteriophage diversity and composition throughout fermentation is critical for elucidating bacteriophage population dynamics and devising more effective strategies to mitigate their impacts⁸³. With the rapid expansion of publicly available phage genome repositories and the continued advancement of viral metagenomics, the taxonomic composition and functional potential of bacteriophages in fermented foods are being progressively uncovered⁸⁴. For instance, Ma et al. employed viral metagenomic approaches to characterize the phage community structure in abnormally fermented vinegar Cupei, thereby providing both compositional and functional insights into the causes of fermentation failure. Strikingly, the dominant viral families in abnormally fermented vinegar mash differed substantially from those in normally fermented counterparts, accompanied by pronounced shifts in community composition and structure. Moreover, abnormally fermented samples exhibited a reduced proportion of lysogenic phages alongside an elevated abundance of bacterial cell wall-degrading enzymes, collectively indicating a greater propensity toward lytic infection of bacterial hosts⁸⁵. Building on these findings, Zhang et al. integrated viral metagenomics with amplicon sequencing to monitor the dynamic succession of phages and microbial communities during strong-flavor Baijiu fermentation, and further quantified phages–microbe associations through correlation analysis. The results revealed that hosts of phages within the Baijiu fermentation system were predominantly non-functional fermentative bacteria, suggesting that viruses may play a regulatory role in shaping microbial community structure and functional potential⁸⁶. Similarly, Ma et al. applied multi-omics strategies to systematically resolve phage and bacterial communities in traditional rice vinegar fermentation. Their study demonstrated that phages and bacterial communities were tightly coupled and responded synchronously throughout the fermentation process, while also exhibiting significant interactions with environmental variables. Notably, phage α-diversity was negatively correlated with bacterial α-diversity, and phage communities displayed continuous contraction and temporal turnover over the course of fermentation⁸⁷. Collectively, these studies establish a theoretical framework for the targeted modulation of phage communities and the competitive dynamics between phages and their bacterial hosts in fermentation microbial ecosystems.

Elucidating bacteriophage–bacteria interaction mechanism provides a novel, integrative framework for understanding the stability and balance of fermentation microbial ecosystems. For example, Fallico et al. used whole-genome microarray analysis to characterize the global molecular response of Lactococcus lactis during early infection by lytic phage c2. Their study identified 61 differentially transcripted genes across 14 functional categories, revealing host regulatory adaptations that mitigate phage-induced membrane perturbations⁸⁸. Moreover, Ledormand et al. examined phage-bacteria interplay in fermented alcoholic beverages. By infecting four Lactobacillaceae strains with virulent phages isolated from cider, they employed integrated transcriptomic and proteomic profiling to capture host post-infection responses. Their results indicated that phage invasion triggered distinct, strain-specific adaptations affecting key functional processes such as cell motility, carbohydrate, amino acid, and nucleotide metabolism⁸⁹. Differently, Jung et al. combined viral and bacterial community profiling with functional annotation of highly abundant viral genes to gain further insights. They demonstrated that during rice wine fermentation, bacteriophages selectively target specific bacterial strains, thus modulate microbial succession. Importantly, beyond their lytic functions, phages were shown to harbor a substantial repertoire of virus-encoded genes beneficial to their hosts. These genes may enhance bacterial adaptability to stress conditions and ultimately improve host survival⁹⁰. Collectively, studies on bacteriophages within fermentation microbial ecosystems have substantially advanced our understanding of microbial community composition and succession, underscoring the potential ecological roles of phages in shaping the microbial consortia of traditional fermented foods.

With the growing recognition of bacteriophages in fermentation microbial ecosystems, an increasing number of studies have begun to explore their ecological roles and functional significance. However, current viromics research remains constrained by incomplete reference databases and limited annotation accuracy, which hampers the reliable interpretation of a substantial proportion of viral sequences and introduces considerable uncertainty in phage–host association predictions. Moreover, predator-prey dynamic models describing phage-host interactions typically rely on high-frequency time-series datasets, whereas most fermentation studies still lack sufficiently high-resolution longitudinal sampling to support such analyses. Furthermore, the industrial application of bacteriophages as biocontrol agents requires careful evaluation of potential risks, including the emergence of phage-resistant hosts and phage-mediated horizontal gene transfer⁹¹. Therefore, future progress will depend on the establishment of standardized analytical pipelines, the development of causal validation strategies, and integrated frameworks combining experimental approaches with computational modeling. Such advances are essential to more reliably translate and phage-driven ecological principles into actionable and industrially implementable control measures.

Multi-dimensional characterization-driven process control framework for standardization

Although multi-dimensional characterization has provided an unprecedented data foundation for deciphering the microbial ecosystems of traditional fermented foods, substantial challenges remain in translating these insights into practical and implementable process control frameworks. In the future, the deeper integration of advanced characterization technologies with systems biology, machine learning, and process engineering is anticipated. This will enable fermentation regulation to gradually shift from a descriptive paradigm toward predictive and controllable modulation, thereby promoting the stabilization and standardization of traditional fermented food production. Importantly, the regulation of fermentation microbiomes is not a single-dimensional optimization problem, but rather a systematic strategy involving multiple critical steps. Accordingly, the following section takes starter optimization, contaminant threshold control, and predictive fermentation modeling as representative examples to illustrate how multi-dimensional characterization outputs can be effectively linked with process control strategies, ultimately forming an industrially applicable framework for microbial ecosystem regulation.

Starter optimization

By integrating multi-omics datasets, key functional microorganisms and metabolic pathway modules responsible for quality formation can be systematically identified, while critical hub taxa within microbial interaction networks can be further elucidated⁹². Moreover, by combining metatranscriptomics, metabolic flux analysis, and absolute quantification approaches, the optimal inoculation ratios of functional strains and their dynamic succession patterns throughout fermentation can be determined. This enables the rational construction of more robust starters or synthetic microbial consortia with enhanced stability and resistance to process perturbations. In addition, microbial ecological fitness (e.g., niche occupation capacity and tolerance to fluctuations in processing parameters) as well as interspecies interactions should be comprehensively considered to facilitate the development of standardized starters with stronger industrial adaptability.

Contaminant threshold control

Through integrated metagenomic and metatranscriptomic analyses, key risk-associated microbial groups and functional genetic markers closely linked to spoilage, toxin production, or fermentation failure can be identified, while metabolomics can further characterize risk-related metabolic fingerprints such as precursor compounds of contaminants or off-flavor metabolites⁹³. On this basis, absolute quantification technologies provide a quantitative foundation for establishing operational contaminant threshold systems, enabling the measurable management of risk microorganisms and their associated metabolites. Furthermore, coupling such threshold systems with real-time sensor monitoring data (e.g., pH, temperature, dissolved oxygen, and CO₂) may facilitate early warning and proactive intervention, thereby preventing severe economic losses caused by fermentation failure.

Predictive fermentation modeling

Machine learning and systems biology models can be employed to analyze time-series multi-omics datasets, extracting key feature variables such as the absolute abundance of core microbial taxa, expression levels of critical functional genes, and concentrations of pivotal metabolites. These variables can then be used to establish predictive models for fermentation outcomes, including final product quality attributes (e.g., acidity, esters, alcohols, and amino acids), flavor profiles, and safety risks. Additionally, incorporating constraints derived from microbial community dynamics and metabolic networks can improve the interpretability and robustness of predictive models in capturing the complex temporal behavior of fermentation microbiomes⁹⁴. Ultimately, predictive models can be mapped onto key process parameters and integrated with real-time monitoring systems to enable closed-loop dynamic control of fermentation, thereby improving batch-to-batch consistency and overall product stability.

Artificial regulation of microbial ecosystems

The comprehensive characterization of fermentation microbial ecosystems ultimately aims to achieve the precise and rational modulation, thereby ensuring product stability and enhancing quality. Drawing on the systematic insights outlined in Section “Multi-dimensional characterization of microbial ecosystems with the development of emerging technologies”, this section synthesizes current artificial control strategies across three interconnected tiers, including microbial community composition, bacteriophage activity, and key fermentation parameters (Fig. 2).

Fig. 2. Artificial regulation of fermentation microbial ecosystems.

a Construction of functional microorganisms to strengthen the fermentation process. b Modulation of fermentation processes using bacteriophages to optimize community structure. c Precise control of physicochemical parameters from raw materials to novel equipment.

Microbial strategies for controlling fermentation microbial ecosystems

A direct and effective approach to improving product quality is the targeted reinforcement of fermentation with functional microorganisms, which are typically isolated, screened, and evaluated from fermentation microbial ecosystems. Such interventions are designed to address three core aspects as follows.

Reducing harmful compounds and improving product quality

Fermented foods such as fermented meat products are rich in proteins and fats, which predispose them to excessive oxidation or hydrolysis during fermentation, negatively affecting product quality⁹⁵. Maintaining a dynamic balance between protein and fat hydrolysis and oxidation is essential. Numerous studies have explored the antioxidant potential of microorganisms derived from fermented meat, including lactic acid bacteria⁹⁶ and yeasts^97,98. Liu et al. applied the selected yeast strains Y12-3 and Y12-4, which exhibited strong antioxidant activity, in combination with Lactobacillus rhamnosus YL-1 in fermented sausages (with L. rhamnosus YL-1 alone as the control), the results showed significant reductions in peroxide value and thiobarbituric acid reactive substances. Moreover, lipid oxidation-related compounds such as 2-hexanal, heptanal, nonanal, and (E)-2-decenal were markedly decreased, demonstrating that these yeast strains can serve as potential antioxidant agents in fermented sausage production²⁴.

Liu et al. applied the selected yeast strains Y12-3 and Y12-4, which exhibited strong antioxidant activity, in combination with Lactobacillus rhamnosus YL-1 in fermented sausages (with L. rhamnosus YL-1 alone as the control). The results showed that the addition of yeast significantly reduced the peroxide value and thiobarbituric acid reactive substances (TBARS) in the fermented samples. Furthermore, the levels of lipid oxidation-related compounds, including 2-hexanal, heptanal, nonanal, and (E)-2-decenal, were markedly decreased. These findings indicate that the selected yeast strains can serve as potential antioxidant starter cultures in fermented sausage production.

Compensating for suboptimal fermentation conditions to improve efficiency and flavor

Soy sauce fermentation involves two sequential stages: koji (solid-state) fermentation and moromi (liquid-state) fermentation, typically requiring an extended production period. Moromi fermentation is conducted in open vats, low ambient temperatures in winter can reduce fermentation efficiency and prolong the fermentation process. Li et al. simulated winter production conditions under laboratory settings and sequentially inoculated Tetragenococcus halophilus at the early stage of moromi fermentation (lactic acid fermentation phase) and Wickerhamomyces anomalus on the fifth day (alcohol fermentation phase), thereby enhancing both lactic acid and ethanol fermentation and promoting ester formation. The approach enabled fermentation completion within 30 days under low temperatures, yielding higher levels of free amino acids and aromatic compounds compared with controls. This strategy offers a practical solution to enhance fermentation efficiency and flavor under cold conditions⁹⁹. Furthermore, previous study demonstrated that using defatted soybeans in soy sauce production improved efficiency and reduces costs¹⁰⁰. However, the lack of triacylglycerols in defatted soybeans limits the production of long-chain and unsaturated short-chain fatty acid esters, which are critical for aroma. Therefore, Zhao et al. isolated two oleaginous yeasts, Sporidiobolus pararoseus and Rhodotorula mucilaginosa, from soy sauce fermentation samples. These strains exhibited robust growth under extreme conditions (high salt, low pH), and could accumulate neutral lipids and continuously release free fatty acids according to lipidomic analyses, thereby promoting ester synthesis. Application of these yeasts in soy sauce fermentation significantly increased long-chain fatty acid esters and key flavor compounds, resulting in a fuller and more harmonious aroma when using defatted soybean substrates¹⁰¹.

Enhancing functional compounds to improve health benefits of fermented foods

Shanxi aged vinegar (SAV) is rich in phenolic compounds such as ferulic acid, which can scavenge reactive oxygen species like hydrogen peroxide and superoxide radicals¹⁰². However, most ferulic acid in SAV exists in bound forms, limiting its antioxidant activity. Aspergillus awamori produces feruloyl esterase, which effectively degrades plant cell walls to release free ferulic acid¹⁰³. Therefore, Zhang et al. inoculated A. awamori to augment the fermentation of SAV. This bioaugmentation strategy elevated the ferulic acid content to 11.31 mg/100 g, marking a 39.4% increase relative to the conventional fermentation group (8.11 mg/100 g). Furthermore, the bioaugmented group exhibited higher concentrations of total polyphenols, total flavonoids, and organic acids, along with a more diverse profile of volatile flavor compounds. Collectively, these findings demonstrate that A. awamori-enhanced fermentation can significantly improve both the functional value and flavor complexity of SAV¹⁰⁴.

Beyond targeted inoculation with defined microbial strains, fermentation can be advanced by leveraging interspecies interactions to biosynthesize key flavor compounds. In soybean paste, phenolic derivatives such as 4-vinylguaiacol (4-VG) and 4-ethylguaiacol (4-EG) provide essential smoky and fruity aroma. However, conventional industrial yeasts like Zygosaccharomyces rouxii cannot convert precursors into these phenols. In contrast, Bacillus licheniformis decarboxylates ferulic acid or phenylalanine to 4-VG via the ferulic acid pathway, while Wickerhamiella versatilis further reduces 4-VG to 4-EG and produces ethyl esters that contribute fruity-sweet aromas^105,106. Therefore, Wang et al. implemented a co-culture of Bacillus subtilis and W. versatilis in soybean paste fermentation. Their findings revealed that W. versatilis effectively converted the 4-VG generated by Bacillus subtilis into 4-EG, leading to a pronounced enhancement of smoky and fruity aroma in the final products. This case demonstrates how targeted microbial interactions, such as metabolic cross-feeding, can be harnessed to direct the synthesis of critical flavor compounds and thereby elevate product quality¹⁰⁷.

Compared with using single or a few functional microorganisms, constructing well-defined synthetic microbial communities (SynComs) aims to emulate and optimize complex natural microbial ecosystems, enabling more stable and controllable fermentation. Beyond reinforcing complex fermentation processes with functional strains, recent studies have focused on constructing and applying SynComs to mimic natural microbial ecosystems for traditional fermented food production. Strategies for constructing these SynComs are generally categorized as bottom-up or top-down approaches.

The bottom-up approach begins with individual pure-culture microorganisms, combining them based on abundance, functionality, or interspecies interactions to establish SynComs. Numerous studies have applied SynComs in fermentation of traditional fermented foods, achieving the desired outcomes (Table 2). Furthermore, the use of well-defined synthetic microbial consortia facilitates a clearer understanding of the microbial ecosystems involved in fermentation. For instance, Wolfe et al. analyzed sequencing data from 137 cheese rind microbiomes across 10 countries, identifying 24 widely distributed bacterial and fungal species as core members. In vitro models of these communities reproduced microbial compositions and successional patterns comparable to those observed in situ¹⁰⁸.

Table 2.

The synthetic microbial communities for the fermentation of traditional fermented foods

Fermentation food	Synthetic microbial communities	Design frame	Results	Further study	Reference
Cheese Rind	Mold: Penicillium, Scopulariopsis, Aspergillus; Yeast: Galactomyces, Debaryomyces, Candida; Bacteria: Corynebacterium, Halomonas, Pseudomonas, Pseudoalteromonas, Vibrio, Staphylococcus, Arthrobacter, Brevibacterium, Brachybacterium, Serratia, Leuconostoc, Yaniella, Nocardiopsis	Identified 24 dominant bacterial and fungal genera were identified from 137 cheeses ↓ Developed an in vitro system to reconstruct rind communities, simulating three rind types (bloomy, natural, washed). Environmental factors (moisture, salinity, pH) were manipulated	In vitro experiments successfully recapitulated the three rind community types. Widespread positive and negative bacterial-fungal interactions were identified.	●Future work should focus on the molecular mechanisms of species interactions, community stability, evolutionary dynamics, and the roles of stochastic versus deterministic forces in community assembly. Genetic tools available for dominant genera can facilitate mechanistic studies.	108
Soy sauce	Mold: Aspergillus oryzae; Yeast: Zygosaccharomyces rouxii; Bacteria:Tetragenococcus halophilus (uninoculated as control, inoculated with T. halophilus, inoculated with Z. rouxii, co-inoculated with T. halophilus and Z. rouxii, inoculated with T. halophilus, followed by sequential inoculation of Z. rouxii when the pH dropped to 5.0)	Selected T. halophilus and Z. rouxii, contributed to the aroma profiles of soy sauce based on literature reports ↓ The mixture of soy flour and wheat flour was inoculated with A.oryzae spores and then transferred into sterile Petri dishes for Koji fermentation ↓ Soy moromi were prepared based on 5 trials (T. halophilus and Z. rouxii), then the mash was homogenized with a vortex and incubated ↓ SPME-GC/MS analysis was used for evaluating the in vitro production of microbial volatile organic compounds	The inoculation sequence (co-inoculation and sequential) has impacts on volatile compound profiles during moromi fermentation. It demonstrated the antagonistic relationship between T. halophilus and Z. rouxii, but it seemed to favor the aroma profiles in the case of sequential inoculation.	●Investigated the inoculum size and ratio between T. halophilus and Z. rouxii in order to fully explore the interaction between the two microbes.	168
Bean Paste	Mold: Aspergillus oryzae; Yeast: Zygosaccharomyces rouxii; Bacteria: Bacillus subtilis, Staphylococcus gallinarum, Weissella confusa	Deconstruct the microbial structure by relative and absolute quantitative analyses ↓ Identified and isolated the representative species from broad bean paste with the highest relative abundance in each corresponding genus and evaluated their fermentation function ↓ Inoculated five core species with initial ratios according to the in situ fermentation into the mixed Pei ↓ Volatile compound analysis by HS-SPME/GC-MS	Realized efficient fermentation of broad bean paste for 6 weeks in a sterile environment with 6% salinity. Microbial community succession and flavor characteristics in in vitro assays highly similar to that of the in situ fermentation process.	●The number of species and the composition, structure, and functionality of synthetic ecosystems should be synthetically optimized in the future	169
Vinegar	Bacteria: Acetobacter pasteurianus, Lactobacillus acetotolerans, Lactobacillus helveticus, Acetilactobacillus jinshanensis	Identified vital and dominant bacterial operational taxonomic units based on metagenomic, metatranscriptomic and 16S rRNA gene sequencings from the autochthonous vinegar starter.Targeted 4 core species. Evaluated the successions and interactions of these 11 bacterial OTUs during the fermentation process ↓ Defined starter was constructed with 4 core species isolated from the autochthonous starter for vinegar fermentation ↓ Compared the effect of autochthonous and defined starters on the AAF process in the condition of sterilized and unsterilized raw materials	The defined starter culture could rapidly initiate the acetic acid fermentation (AAF) in a sterile or unsterilized environment, and similar dynamics of metabolites and environmental indexes of fermentation were observed as compared with that of autochthonous starter (P > 0.05).	●The four strains selected were only the representative strains from the autochthonous starters so the other seven OTUs should be evaluated. ●The species proportion and inoculation method of defined starter can be optimized to obtain better fermentation effects	170
Strong-flavor Baijiu	Mold: Aspergillus niger, Monascus purpureus; Yeast: Saccharomyces cerevisiae (3 strains); Bacteria: Burkholderia sp., Ligilactobacillus acidipiscis, Clostridium beijerinckii, Clostridium tyrobutyricum,Clostridium butyricum, Caproiciproducens sp., Clostridium kluyveri	Design microbial combination based on function characteristics ↓ Isolated strains with strong ability and constructed functional microbial combination ↓ Simulated solid fermentation of Strong-Flavor Baijiu under controlled temperature(not exactly sterile conditions) ↓ The volatile compounds in the fermented grains were assayed by HS-SPME-GC-MS	The combination of functional microorganisms showed a good performance in raw material consumption and flavor chemical synthesis of Baijiu	●The number of species and the composition, structure, and functionality of synthetic ecosystems should be synthetically optimized in the future	53
Fermented sausage	Lactobacillus plantarum LZ-4, Staphylococcus saprophyticus SZ-7, Macrococcus caseolyticus SM-2, Streptococcus thermophilus SL-17, Pediococcus pentosaceus LP-1	Selected five dominant genera based on flavor-producing ability, co-occurrence analysis, and relative abundance ↓ Identified one species(representative strains) with the highest relative abundance in each corresponding genus ↓ Sausages were fermented under controlled conditions(temperature and relative humidity) ↓ The volatile compounds were measured by HS-GC-MS and sensory analysis by 11 systematically trained evaluators	The synthetic microbial community could reproduce some key volatile flavor compounds and showed better performance in aroma profile than commercial starter cultures.	●Optimize the selection principles identifying the core microbial community of spontaneously fermented sausages, to build a more representative synthetic microbial community	171
Paocai (a traditional fermented vegetable)	Yeast: Kazachstania exigua; Bacteria: Liquorilactobacillus nagelii, Pediococcus ethanolidurans; Lentilactobacillus buchneri	Used metagenomic sequencing to analyze the microbial community structure of Paocai brine during fermentation, combined with SBSE-GC-O/MS to analyze volatile flavor compounds ↓ Screened keystone microbiota in key flavor formation through abundance analysis, co-occurrence network analysis, redundancy analysis (RDA), and Spearman correlation analysis ↓ Isolated target strains and constructed a synthetic keystone microbiota for inoculation fermentation validation	Paocai fermented with the synthetic keystone microbiota exhibited volatile organic compound profiles highly similar to those fermented with reused brine.	●The mechanisms of microbial interactions within the community remain unclear, further study is needed for practical industrial application of synthetic keystone microbiota for stable quality production	14
Xiaoqu light-aroma Baijiu	Mold: Rhizopus; Yeast: Pichia, Wickerhamomyces, Issatchenkia, Saccharomyces, Kazachstania; Bacteria: Acetobacter, Bacillus, Lactobacillus, Weissella	Identified core microbiota based on microbial abundance, occurrence frequency, co-occurrence networks, and correlations with flavor compounds ↓ A synthetic microbial community was constructed and validated through simulated fermentation using 10 representative species (bran Qu preparation) ↓ Tolerance assays and microbial interaction experiments were conducted to elucidate the succession mechanisms of core microbes.	The ten identified core microbial genera were able to reproduce the fermentation process and flavor characteristics of Xiaoqu light-aroma Baijiu.	●Future research should focus on the impact of environmental disturbances on the assembly and succession of core microbiota, and construct a series of synthetic microbial communities adapted to different environmental conditions to achieve stable production under different conditions and in different regions.	172

In contrast, the top-down approach starts with a complex natural microbial community and gradually enriches target functional groups through selective cultivation, making it particularly suitable for microorganisms that are difficult to isolate^109,110. For example, ester precursors in strong-flavor Baijiu fermentation are primarily derived from pit mud located at the bottom of fermentation pits. Effective interaction between the lower-layer mash and pit mud microorganisms facilitates high-quality base liquor production. However, the upper-layer mash has limited contact with pit mud, resulting in lower concentrations of flavor compounds, particularly esters, and thereby compromising base liquor quality¹¹¹. However, the anaerobic preference and culturability of pit mud microbiota complicate the construction of synthetic communities via bottom-up approaches. To address this, Gao et al. adopted a top-down strategy by selectively enriching the pit mud microbiota using lactic acid (a major component of fermentation yellow water) as the carbon source. This yielded a simplified, Caproiciproducens-dominated community through successive passaging, which notably avoided the production of off-flavor compounds like p-cresol and phenol. When applied to the upper-layer fermentation mash, the enriched community significantly elevated the concentrations of key flavor esters, such as ethyl hexanoate, ethyl pentanoate, and ethyl octanoate, in the resulting strong-flavor baijiu. This study demonstrates that top-down derived simplified microbial consortia can effectively enhance the sensory quality and flavor profile of fermented foods¹¹².

Regulation of fermentation microbial ecosystems by bacteriophages

Bacteriophages can serve as precise “biological tools” for targeting harmful bacteria due to their high host specificity, offering a unique and powerful strategy to regulate fermentation microbial ecosystems. Consequently, phages have attracted growing scientific interest both as alternatives to conventional antimicrobials for pathogen control and as modulators of microbial community assembly^113,114. Notably, several phages have been approved by the U.S. Food and Drug Administration as “Generally Recognized as Safe”, and phage-based biocontrol has been successfully applied in a variety of foods to eliminate pathogenic bacteria, including milk, cherries, cabbage, and poultry^115,116.

Given the complexity of fermented food systems, phage-mediated modulation of specific strains within mixed microbial communities may have profound effects on fermentation. Therefore, careful evaluation of potential impacts on non-target microorganisms is essential prior to the application of phages in mixed fermentation systems⁸¹. For instance, Tomat et al. isolated Escherichia coli phages DT1 and DT6, which effectively reduced pathogenic and Shiga toxin-producing E. coli during milk fermentation without compromising the growth or fermentation activity of Streptococcus thermophilus. Environmental parameters such as temperature, pH, and metal ion concentrations must be optimized to maintain phage efficacy under dynamic fermentation conditions. In another study, Bandara et al. demonstrated the targeted potential of bacteriophages using two Bacillus cereus-specific phages, BCP1-1 and BCP8-2, in a model cheonggukjang (Korean fermented soybean) fermentation. Phage treatment selectively inhibited Bacillus cereus while allowing the preferred Bacillus subtilis to proliferate. Notably, the presence of divalent cations further enhanced phage lytic activity, highlighting the role of environmental factors in phage-based biocontrol and underscoring the potential of bacteriophages as precise antimicrobial agents in fermentation¹¹⁷.

Bacteriophages can contribute to improving the overall quality of fermented foods¹¹⁸. Lin et al. isolated three Staphylococcus-specific phages from kimchi and investigated their effects on the fermentation process. Bacteriophage supplementation significantly increased the relative abundance of lactic acid bacteria while reducing multiple spoilage-associated taxa, thereby optimizing the microbial community structure. Moreover, bacteriophages treatment accelerated acidification, reduced nitrite accumulation, and increased lactic and acetic acid production, collectively enhancing kimchi quality¹¹⁹. Similarly, Zheng et al. applied two lytic phages targeting dominant early-stage spoilage bacteria in combination with Lactobacillus plantarum during cucumber pickle fermentation. This strategy effectively controlled Enterobacter cloacae and Pseudomonas fluorescens, suppressed nitrite formation, and demonstrated considerable promise for producing high-quality pickles under low-salt fermentation conditions²⁵.

Physicochemical parameter optimization to shape fermentation microbial ecosystems

In addition to direct manipulation of microbial composition, optimization of physicochemical parameters represents a fundamental and industrially scalable strategy for indirectly shaping fermentation microbial ecosystems.

Initial conditions play a decisive role in determining fermentation trajectories, and precise control within optimal ranges can steer the process toward desirable microbial succession and product quality¹²⁰. Understanding the impact of key initial physicochemical parameters on fermentation processes and product quality is therefore critical across diverse fermented foods. As an example, Wang et al. systematically examined how varying initial moisture levels (50–70%) influenced microbial succession and metabolite profiles during Xiaoqu baijiu fermentation. Higher moisture content accelerated the establishment of lactic acid bacteria-dominated communities and reinforced the stability of microbial interaction networks. Notably, fermentation initiated at 70% moisture yielded the highest total volatile content and an increased ethyl acetate to ethyl lactate ratio, contributing to a more desirable flavor profile. This illustrates how targeted modulation of initial conditions can directly steer microbial ecosystems and enhance the sensory quality of fermented products¹²¹. Similarly, Yang et al. examined the effects of different initial salt concentrations (0.5–3.5%) on Northeast Chinese sauerkraut fermentation. A low salt level of 0.5% facilitated more complete glucose and fructose utilization, leading to enhanced accumulation of organic acids, including lactic, citric, and succinic acids, among them lactic acid reaching 10.78 g/L by day 30 (p < 0.05). This condition accelerated fermentation maturation and improved both microbial metabolic activity and sensory attributes¹²². In another study, Zhang et al. assessed the impact of initial temperature (28 °C vs 37 °C) on Jiang-flavor baijiu fermentation. The higher initial temperature increased the relative abundance of Lactobacillus, accelerated microbial succession, strengthened interspecies interactions, elevated total acidity and reduced total ester content, which providing practical guidance for temperature regulation in baijiu production¹²³.

Beyond initial settings, dynamic control of fermentation parameters throughout the process is equally critical. Among these factors, temperature is particularly advantageous due to its ease of monitoring and adjustment without disturbing the fermentation microbial system, making it a central focus of traditional fermentation management. In doubanjiang production, fermentation at 40 °C favors the accumulation of amino acid nitrogen and other metabolites, however, prolonged exposure to high temperatures during later stages can lead to excessive acidity and increased biogenic amine formation. Conversely, fermentation at 30 °C effectively limits biogenic amine accumulation but slows the generation of flavor compounds¹²⁴. To reconcile these trade-offs, Niu et al. proposed a two-stage temperature control strategy consisting of an early high-temperature phase (40 °C for 7 days) followed by a lower-temperature phase (30 °C for 49 days). Compared with conventional fermentation conducted at fluctuating temperatures of 20–35 °C for 56 days, this approach promoted rapid early accumulation of amino acid nitrogen, free amino acids, and organic acids while maintaining acceptable total acidity and biogenic amine levels during later stages. As a result, the final doubanjiang exhibited pronounced sauce-like and fruity aromas, demonstrating a practical and efficient strategy for industrial-scale production¹²⁵.

Advances in fermentation equipment further enable precise parameter regulation. Traditional doubanjiang fermentation in ceramic jars is often limited by inefficient heat and mass transfer, sluggish biochemical reactions, and extended fermentation cycles¹²⁶. To address these challenges, Sun et al. developed a closed rotary-drum bioreactor operated under dynamic temperature conditions for doubanjiang-meju fermentation. This system enhances mass transfer efficiency and accommodates the diverse temperature requirements of microbial growth and enzymatic activity. Comparative analyses showed that products obtained from the rotary-drum fermenter possessed superior physicochemical properties, with volatile compound contents 27.24% and 130.72% higher than those produced under constant-temperature and traditional fermentation conditions, respectively, along with the lowest detected levels of aflatoxin B1. Sensory evaluation further revealed a more desirable red-brown color, intensified sauce flavor, prolonged fragrance retention¹²⁷. Collectively, these findings underscore the importance of optimizing fermentation parameters and equipment to achieve high-quality fermented foods production.

Regulatory, biosafety, and industrial trade-offs of artificial regulation

Notably, although artificial regulation strategies—including the enhancement of functional strains, the introduction of synthetic microbial communities, and bacteriophage-mediated biocontrol—have provided promising avenues for stabilizing the production of traditional fermented foods, their industrial translation is accompanied by substantial regulatory and biosafety challenges. (1) The introduction of functional strains or synthetic microbial communities must comply with food safety regulations across different countries and regions, such as the widely recognized GRAS (Generally Recognized as Safe) or QPS (Qualified Presumption of Safety) assessment frameworks, and requires systematic strain traceability, whole-genome safety evaluation, and screening for antibiotic resistance genes and virulence factors. (2) High-density fermentation environments may facilitate horizontal gene transfer among microorganisms, elevating the risk of dissemination of antibiotic resistance genes or mobile genetic elements within fermentation systems and throughout the food chain. (3) Caution is warranted in the application of bacteriophage-based regulation, with critical attention to the evolution of bacteriophage resistance, transduction events mediated by lysogenic bacteriophages, and potential disturbances to the production environment microecology caused by bacteriophage release, so as to guarantee controllability and traceability in industrial practice. (4) Industrial implementation of artificial regulation strategies also entails trade-offs between cost and sustainability, for instance, online monitoring systems and precise process control parameters substantially increase energy consumption and capital investment, while cleaning protocols for fermentation equipment and the disposal of fermentation by-products impose stricter demands on environmental sustainability. (5) Uncertainties remain regarding consumer acceptance of artificially modulated fermentation processes, as well as corresponding variations in product flavor profiles and quality attributes. Transparent risk communication and standardized management are therefore equally essential for facilitating industrial-scale production. In summary, alongside the advancement of artificially regulated fermentative microbial ecosystems, there is an urgent need to establish a comprehensive evaluation framework that integrates food safety, regulatory compliance, economic viability, and environmental sustainability. This will enable the industrialized, standardized, and green development of traditional fermented food manufacturing.

Challenges from empirical practice to a scientifically controlled fermentation

Advances in modern analytical techniques have enabled systematic characterization of the microbial ecosystems in traditional fermented foods. Leveraging this knowledge, targeted modulation of fermentation microbial ecosystems can steer the process toward desirable outcomes, mitigating fermentation failures and enhancing product quality. Nonetheless, the complete transition of traditional fermented food production from empirical practice to a scientifically controlled process remains a formidable challenge (Fig. 3).

Fig. 3. The future challenges from empirical practice to a scientifically controlled fermentation of traditional fermented foods.

a Unlocking uncultured microorganisms. b The systematic design and construction of SynComs. c The development of intelligent fermentation systems.

Uncultured microorganisms in fermentation microbial ecosystems

Currently, the isolation of functional microorganisms in fermentation samples largely relies on traditional dilution plating. However, many low-abundance or uncultivable microorganisms within fermentation microbial ecosystems may possess critical functional roles. Factors such as nutrient dependency, dormancy, symbiotic relationships, and technical limitations have hindered functional metabolic studies of these organisms, highlighting the need for innovative cultivation strategies¹²⁸.

Approaches developed for complex microbiomes, including soil, seawater, and the human gut, offering valuable insights for isolating rare or previously uncultivable microorganisms from fermented foods or searching for the key functional enzymes^129,130. Targeted methods, such as reverse genomics, live-cell FISH, and Raman-activated cell sorting, as well as high-throughput cultivation techniques like membrane diffusion, droplet microfluidics, and nanoculture, have proven effective for microbial sorting and isolation^130–135. Notably, He et al. developed the FISH-scRACS-seq (Fluorescence-in situ-hybridization-guided single-cell Raman-activated sorting and sequencing) approach and successfully applied it to explore functional microbial resources in seawater and soil microbiomes. In this method, FISH is employed to target specific microbial populations, while scRACS-seq assesses metabolic activity and isolates functionally active single cells via a microfluidic sorting chip RAGE (Raman-activated gravity-driven encapsulation) for subsequent whole-genome sequencing. This technology enables direct linkage of microbial taxonomy, function, and genome at the single-cell level¹³⁶. In the future, by integrating traditional and emerging cultivation techniques with multi-omics analyses, researchers can strategically sample key spatiotemporal stages of fermentation, isolate functional strains, characterize their metabolic activity, and sequence their genomes, thereby expanding the repository of functional microorganisms for fermented foods.

However, despite the considerable application potential of these advanced cultivation and isolation technologies, multiple practical constraints still limit their implementation in industrial fermentation microbiome research and strain development. First, most of these cutting-edge approaches remain in the early to mid-stages of technological maturation, and their operation typically relies on highly specialized instruments, microfluidic chips, and skilled personnel. This substantially increases the initial investment required for industrial adoption, while the overall cost-benefit ratio and economic return remain uncertain. Moreover, although high-throughput cultivation and sorting technologies are theoretically capable of processing large microbial populations, their throughput is often compromised in practice due to workflow complexity, limited system robustness, and challenges associated with post-isolation microbial recovery and cultivation, thereby weakening their claimed high-throughput advantages. In addition, the compatibility of these technologies with traditional fermented food systems is not always reliable. The complex matrices of fermentation samples, often characterized by high salinity, low pH, and abundant organic compounds, may interfere with high-throughput data acquisition and sorting efficiency. Furthermore, even when microorganisms are successfully isolated, challenges remain in ensuring that they retain metabolic activity and exert their authentic functional roles after being reintroduced into their native fermentation ecosystems. Therefore, future progress will require not only continuous technological innovation but also the establishment of standardized and cost-controllable operational protocols. Equally important is the implementation of systematic benchmarking and validation across diverse fermentation systems, along with comprehensive performance evaluation of isolated strains under industrially relevant fermentation conditions.

The design and construction of SynComs

Many studies have leveraged SynComs at the laboratory scale to regulate or reconstruct traditional fermentation microbial ecosystems. However, the design and construction methods of these SynComs vary considerably, making it challenging to systematically summarize and compare results across different studies. Consequently, establishing standardized approaches for the systematic design and construction of SynComs is of critical importance.

Regarding SynComs design, most current studies rely on multi-omics analyses to identify key functional microorganisms through correlation-based methods. While these approaches provide valuable insights, the sheer volume of omics data and the limitations of analytical methods can hinder accurate identification of truly critical functional microorganisms. Machine learning (ML) offers a powerful alternative, as it can integrate observations across multiple scales, extract meaningful patterns from massive datasets, and establish predictive frameworks^8,137. In the context of traditional fermented foods, ML has already been applied, for example, to predict pH changes during cheese fermentation, thereby forecasting fermentation endpoints and optimizing downstream operations¹³⁸. It has also been used to predict product quality and fermentation progress^139–142. However, the application of machine learning (ML) to the design of Synthetic Microbial Communities (SynComs) remains a largely underexplored frontier. Going forward, ML holds significant potential for mining multi-omics datasets to identify key microorganisms and fermentation parameters that shape microbial ecosystems, thereby enabling the iterative optimization of SynComs within a “design–build–test–learn” synthetic biology cycle. Furthermore, the strategic incorporation of microorganisms capable of synthesizing functional components—such as B vitamins, bioactive peptides, or γ-aminobutyric acid—could augment the nutritional profile of traditional fermented foods. Equally critical is ensuring the stability of SynComs during fermentation. For instance, Tan et al. illustrated that metabolic stability in Baijiu fermentation is governed by fermentation parameters, dynamic resource allocation between yeasts and Lactobacilli, and inherent metabolic redundancy¹⁴³. Standardizing the design process, which requires the systematic incorporation of multiple factors, is therefore crucial to advance a rigorous research paradigm and enable practical implementation.

Regarding SynComs construction, most studies manually culture pure strains and mix them in predetermined ratios, this is both labor-intensive and prone to variability. Automated experimental platforms, such as liquid-handling robotic manipulation, can reduce labor demands while improving precision and diversity in microbial combinations¹⁴⁴. In addition, emerging technologies also hold promise for constructing SynComs for traditional fermented foods. For example, 3D bioprinting has been applied to yogurt fermentation. Lactobacillus bulgaricus and Streptococcus thermophilus were loaded into separate bioinks and alternately deposited within a hydrogel lattice using 3D printing, creating a spatially segregated composite starter. This spatially structured hydrogel of SynComs maintained consistent fermentation performance over 25 consecutive cycles and exhibited greater stability compared to conventional liquid co-culture systems¹⁴⁵.

However, machine learning-assisted design of SynComs is theoretically more forward-looking and rational, its practical implementation is still substantially constrained by omics data quality and limited model interpretability. First, machine learning-based analyses of fermentation datasets are highly susceptible to overfitting, batch effects, and poor cross-scenario generalization. Variations in raw material sources, seasonal fluctuations, geographic origins, and factory-specific processing conditions may all lead to inaccurate predictions or even model failure. Therefore, to achieve meaningful industrial value, machine learning-guided SynComs design requires rigorous causal validation and robust multi-batch industrial datasets to ensure reliability and transferability, which inevitably increases the cost and resource investment prior to industrial deployment. Second, compared with laboratory-scale settings, maintaining stable community composition and preserving functional performance under industrial fermentation conditions remains highly challenging. Moreover, the ecological stability and long-term resilience of SynComs within complex industrial fermentation ecosystems are still largely uncertain. Consequently, although emerging technologies such as machine learning-driven SynComs design, automated consortium construction platforms, and 3D bioprinting represent important future directions, their industrial applicability will ultimately depend on whether key bottlenecks—including cost control, process standardization, validation under real production environments, and enhanced ecological robustness—can be effectively addressed. Future efforts should therefore seek a balance between technological innovation and engineering feasibility, integrating advanced computational approaches with scalable fermentation engineering strategies to facilitate the transition of SynComs from proof-of-concept studies toward industrially implementable applications.

Real-time monitoring and automated control of fermentation microbial ecosystems

Traditional fermentation monitoring is predominantly dependent on manual expertise and intermittent sampling coupled with complex laboratory analyses. This approach introduces uncertainties and temporal delays in obtaining results, which hinders timely guidance for process control. Furthermore, intermittent sampling is labor-intensive, risks disturbing the indigenous microbial ecosystem, and fails to ensure consistent product quality. Consequently, there is a growing imperative for non-invasive, real-time monitoring and dynamic regulation of fermentation microbial ecosystems. The development of integrated fermentation equipment, capable of real-time acquisition and control of key physicochemical parameters, enables rapid and continuous process monitoring and adjustment without disrupting the fermentation process.

The Internet of Things (IoT) have the capability of integrating sensors with cloud computing platforms, allowing real-time data collection, upload, analysis, and visualization, thereby providing online guidance for dynamic fermentation control¹⁴⁶. For example, Sotirios et al. developed a SmartBarrel system based on IoT (Internet of things) sensor technology to monitor and predict grape wine fermentation. This system incorporates an electronic nose and electronic tongue. The electronic nose tracks gaseous emissions during fermentation, while the electronic tongue monitors changes in acidity, sugar content, and color. The collected data are transmitted to a cloud-based visualization platform and analyzed using a novel V-LSTM (Variable-length Long Short-Term Memory) deep learning model for intelligent predictive monitoring¹⁴⁷. In addition to monitoring physicochemical parameters, real-time tracking of microbial community dynamics is equally essential. Various sensors have been developed for this purpose, including optical density sensors to track turbidity in grape wine fermentation, and fluorescence spectroscopy to assess yeast metabolic activity in tempeh¹⁴⁸. Looking forward, the integration of genomic and multi-omics approaches could enable real-time mapping of fermentation microbiomes. Beyond monitoring and prediction, precise control of fermentation is crucial. Adaptive control strategies can dynamically regulate the process in response to changing conditions, making them highly advantageous for traditional food fermentation¹⁴⁹. Therefore, future developments can focus on intelligent fermentation systems tailored for traditional foods, integrating biosensors, real-time monitoring, machine learning, and automated control to achieve standardized and high-quality production.

Although intelligent fermentation systems incorporating real-time monitoring and automated control technologies are highly attractive for the future industrial production of traditional fermented foods, their large-scale implementation still faces multiple practical constraints. First, traditional fermented food manufacturing is often characterized by highly heterogeneous solid-state or semi-solid substrates, which makes the deployment, sterilization, and long-term stable operation of in situ sensors considerably more challenging. Moreover, complex fermentation matrices frequently exhibit high viscosity, elevated particulate content, and biofilm formation, all of which can cause sensor fouling, signal drift, and measurement interference, thereby substantially compromising the accuracy and reliability of monitoring data. Second, real-time monitoring of fermentation microbiota remains at an early stage of technological development. Although indirect sensing approaches, such as optical density measurement and fluorescence spectroscopy, can rapidly provide dynamic signals, they generally lack sufficient taxonomic resolution and therefore cannot precisely capture changes in microbial community composition. Furthermore, integrating genomics-based and multi-omics approaches into real-time monitoring systems is still hindered by long detection cycles and limited automation, indicating that true online and real-time microbial surveillance remains far from being fully achievable. In addition, automated control systems require clearly defined control objectives, response thresholds, and robust causal linkages between sensor signals and fermentation outcomes. However, such fundamental mechanistic relationships still demand extensive systematic investigation and rigorous validation. Only by addressing these practical bottlenecks can intelligent fermentation systems become scalable, economically feasible, and widely applicable solutions for the industrialization of traditional fermented food production.

Conclusion

Traditional fermented foods hold a central place in human diets, and fermentation microbial ecosystems playing a crucial role in determining flavor, quality, and product stability. This review systematically summarizes research on the microbial ecosystems of traditional fermented foods, tracing the trajectory from comprehensive multi-dimensional characterization to artificial regulation. Multi-omics technologies have enabled unprecedented insights into this once opaque “black box”, while functional microorganisms, synthetic microbial communities, bacteriophages, and fermentation parameter modulation collectively form a versatile toolkit bridging understanding and practical application.

Advances in molecular biology and high-throughput sequencing have profoundly advanced our understanding of fermentation microbial ecosystems. A suite of “omics” technologies, including 16S/ITS amplicon sequencing, metagenomics, metatranscriptomics, proteomics, metabolomics, and flavoromics, are now routinely employed to decipher microbial composition, functional potential, metabolic networks, community dynamics, and interspecies interactions. Building on this foundational knowledge, targeted microbiome manipulation is being explored to enhance the consistency, quality, and safety of traditional fermented foods. Strategies range from harnessing native microbial functions and interactions, constructing synthetic communities, and applying bacteriophages, modulating physicochemical parameters for improved process control and product standardization.

Despite these advances, translating omics-derived knowledge into predictable, controllable fermentation remains challenging. Future progress hinges on three key directions, including (i) unlocking the functional potential of previously uncultured microorganisms, (ii) establishing rational, systematic design principles for synthetic microbial community, and (iii) developing intelligent fermentation platforms capable of real-time monitoring and dynamic regulation. Achieving these goals will require an interdisciplinary integration of food science, microbiology, synthetic biology, and data science, collectively enabling the standardization, industrialization, and optimization of traditional fermented food production while preserving its rich sensory and cultural heritage.

Author contributions

Mengqin Wu, Data curation, Formal analysis, Writing—original draft; Jiayao Wang, Data curation, Formal analysis; Yu Wang, Data curation, Formal analysis; Youqiang Xu, Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Software, Writing—review & editing; Baoguo Sun, Writing—review & editing.

Data availability

All the data generated for this study are available on request to the corresponding author.

Competing interests

The authors declare no competing interests.