Primary and Secondary Succession Mediate the Accumulation of Biogenic Amines during Industrial Semidry Chinese Rice Wine Fermentation

Understanding the shifting patterns of substance usage and microbial interactions is a fundamental objective within microbiology and ecology. Analyses of primary and secondary microbial succession allow for determinations of taxonomic diversity, community traits, and functional transformations over time or after a disturbance. The kinetics of BA generation and the patterns of resource consumption, functional metagenome prediction, and microbial interactions were profiled to elucidate the equilibrium mechanism of microbial systems. Secondary succession after a disturbance triggers a change in resource usage, which in turn affects primary succession and metabolism. In this study, the functional potential of exogenous microorganisms under disturbance synergized with secondary succession strategies, including rebalancing and dormancy, which ultimately reduced BA accumulation. Thus, this succession system could facilitate the settling of essential issues with respect to microbial traits that rely on resource usage and microbial interactions that occur in natural ecosystems.

ABSTRACT

The use of exogenous functional microorganisms to regulate biogenic amine (BA) content is a common approach in fermentation systems. Here, to better understand the microbial traits of succession trajectories in resource-based and biotic interference systems, the BA-related primary and secondary succession were tracked during industrial semidry Chinese rice wine (CRW) fermentation. Dominant abundance and BA-associated microbial functionality based on phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) indicated that Citrobacter, Acinetobacter, Lactobacillus, Exiguobacterium, Bacillus, Pseudomonas, and Enterobacter spp. prominently contributed to the decarboxylase gene family in CRW. The expression levels of tyrosine decarboxylase (tyrDC), ornithine decarboxylase (odc), and agmatine deiminase (aguA) genes were assessed by quantitative PCR (qPCR). The transcription levels of these genes did not correlate with the BA formation rate during postfermentation, indicating that acidification and carbon source depletion upregulated the expression and microbes launch the dormancy strategy to respond to unfavorable conditions. Furthermore, microbial interference with CRW fermentation by Lactobacillus plantarum (ACBC271) and Staphylococcus xylosus (CGMCC1.8382) coinoculated at a ratio of 1:2 exhibited the best synergetic control of BA content. Spearman correlations revealed that Lactobacillus and Staphylococcus exhibited influence on BA-associated microbiota (|ρ| > 0), Exiguobacterium and Pseudomonas were strongly suppressed by Lactobacillus (ρ = −0.867 and ρ = −0.782, respectively; P < 0.05), and Staphylococcus showed the strongest inhibitory effect toward Lactobacillus (ρ = −0.115) and Citrobacter (ρ = −0.188) in the coinoculated 1:2 group. The high inhibitory effect of exogenous added strains on specific bacteria presented evidence for the obtained BA-associated contributors. Overall, this work provides important insight into the microbial traits that rely on resource usage and functional microbiota within food microbial ecology.

IMPORTANCE Understanding the shifting patterns of substance usage and microbial interactions is a fundamental objective within microbiology and ecology. Analyses of primary and secondary microbial succession allow for determinations of taxonomic diversity, community traits, and functional transformations over time or after a disturbance. The kinetics of BA generation and the patterns of resource consumption, functional metagenome prediction, and microbial interactions were profiled to elucidate the equilibrium mechanism of microbial systems. Secondary succession after a disturbance triggers a change in resource usage, which in turn affects primary succession and metabolism. In this study, the functional potential of exogenous microorganisms under disturbance synergized with secondary succession strategies, including rebalancing and dormancy, which ultimately reduced BA accumulation. Thus, this succession system could facilitate the settling of essential issues with respect to microbial traits that rely on resource usage and microbial interactions that occur in natural ecosystems.

INTRODUCTION

Chinese rice wine (CRW) is a semisolid multispecies fermented alcoholic beverage made from protein-rich rice and starter cultures, which possess abundant amino acids, oligosaccharides, and trace elements (1). Based on the sugar content, CRW is sorted into sweet, semisweet, semidry, and dry types (2), with semidry CRW dominating the market due to its unique characteristics. The open, spontaneous fermentation system and nitrogen-rich source of CRW result in total biogenic amine (BA) levels ranging from trace amounts up to 241 mg/liter (3). The most frequently observed BAs in CRW are putrescine (PUT), tyramine (TYR), histamine, cadaverine, spermine, spermidine, phenylethylamine, and tryptamine. Among these, PUT is typically the most abundant BA, followed by TYR. A small quantity of BA can be metabolized into other physiological substances in the human intestines by amine oxidases (monoamine and diamine oxidases) (4). However, given the differences in the toxicological thresholds to individuals and ethanol being an amine oxidase inhibitor that blocks self-detoxification, high levels of BAs are involved in the systemic circulation and release of adrenaline and norepinephrine, which results in gastric acid secretion, migraine, tachycardia, and elevated blood sugar levels and blood pressure (5).

BA formation is dependent on the presence of precursor amino acids and microbial decarboxylases, as well as an appropriate environment (6). The accumulation of BAs is a strategy leveraged to protect acid-resistant mediators or to obtain metabolic energy by coupling amino acid decarboxylation with electrically charged amino acid reverse transport proteins (7). BA formation counteracts intracellular acidification, which increases survival under acidic stress conditions and helps to restore the internal pH through proton depletion and the excretion of BAs and CO₂. Thus, BA accumulation represents a cellular defense mechanism used to resist acid stress. Furthermore, it has been proven that low pH may trigger the transcription of decarboxylase genes (8), which can support primary metabolism under nutrient-poor conditions (9). Lactic acid bacteria are considered a core microbe in BA metabolism (5). However, due to the strain specificity of BAs and horizontal gene transfer between microorganisms, BA-related microbes are widely distributed and include Enterococcus, Lactobacillus, Lactococcus, Leuconostoc, Streptococcus, Pediococcus, and Weissella spp. as minor producers (10, 11). Generally, the previous research on BAs mainly followed the bottom-up design (12, 13). Relevant strains are screened to verify the amine-producing or amine degradation capacity and then to quantify the expression and transcription levels of the decarboxylase or amine oxidase gene to reveal the metabolism of BAs (12–14). This approach can reveal the features of a single strain and result in safe and effective starter cultures. However, within the complicated fermentation system, the strains that can be screened rely on culturable characteristics that cannot lend insights into the fundamental principles of characterization in the microbial ecosystem. Additionally, dominant microbiota do not always exhibit effective targeted metabolic function (15). Therefore, combination of dominant communities with a functional prediction of microbiome systems can provide a good starting point for the isolation of crucial members responsible for BAs by use of bottom-up approaches. Phylogenetic investigation of communities by reconstruction of unobserved states (PICRUSt) is an algorithm by which metagenomes are predicted and annotated based on the 16S rRNA gene sequence data and reference genome databases to obtain functional profiles containing culturable and unculturable microbial communities (16). PICRUSt is widely used in predicting the functionality of gastric (17, 18) and environmental (19, 20) microbial communities as well as contributions of a gene family to a specific type of metabolism.

A broader knowledge of microbial succession is important for predicting the impacts of environmental change on microbe-mediated BA formation in industrial CRW fermentation. Microbial succession has been delineated into primary and secondary succession, terminology originating from plant ecology (21). Primary succession is a type of biological succession in native bare systems in which organisms have never grown or have grown but were wiped out, which is dictated by changes in resources. Unlike resource-based succession, microbial secondary succession occurs following a disturbance to a previously established ecosystem (22). Across microbial ecology, immigrants that govern the indigenous microbiota determine the secondary successional patterns after disturbance (23). Native microbial consortia that persist in disturbed ecosystems and perform metabolic interactions in coexisting taxa can affect community function. Therefore, new community structures may contribute substantially to secondary succession. Employing starter cultures has been demonstrated as a means of regulating BAs in fermented foods, and they cause fewer adverse organoleptic and unhealthy alterations (13, 24). The interference of exogenous strains with the native food fermentation ecosystem triggers secondary succession. However, how changes in the microbial community structure affect potential functions remains unknown. Linking succession strategies and microbial traits to changes in BA accumulation will allow the mechanisms associated with BA formation over primary and secondary succession to be disentangled.

In this study, we depict the in situ characterization and scale of disturbance of microbial communities during industrial CRW fermentation. High-throughput sequencing was performed to assess the microbial diversity, and predictive functional profiling for BA formation was obtained with PICRUSt to reveal the temporal patterns of BA-associated communities. The mechanism of BA formation during the CRW fermentation process was elucidated by quantifying the expression of decarboxylase genes. Furthermore, mixtures of strains in different proportions were inoculated to regulate the dynamics of in situ community assembly to assess how these patterns affect BA accumulation under secondary succession. Overall, the results of our study of this system have the potential to combine in situ community analyses of microbial diversity disturbances to better understand the patterns and underlying mechanisms of community assembly and microbial interactions.

RESULTS

BAs associated with primary succession during CRW fermentation.

Microbial community succession provides a framework for understanding the temporal and spatial changes that occur during microbial fermentation. The results of our study revealed the bacterial and fungal diversity present during CRW fermentation based on Illumina MiSeq sequencing. In the sampled communities, a total of 677 bacterial genera and 77 fungal genera were identified. Only 13 bacterial and 2 fungal genera were observed at >1% average abundances (Fig. 1A), with variable abundances observed at each stage. At the genus level (Fig. 1B), Staphylococcus (0 to 30.44%), Lactobacillus (0.59 to 65.17%), Lactococcus (1.18 to 32.08%), Leuconostoc (2.21 to 20.61%), Acinetobacter (0.08 to 27.55%), Bacillus (0 to 11.68%), Exiguobacterium (0.17 to 36.54%), Citrobacter (0.12 to 23.17%), Oceanobacillus (0 to 6.38%), Pseudomonas (0.05 to 8.93%), Weissella (0 to 2.21%), Kurthia (0.01 to 2.33%), and Enterobacter (0 to 2.65%) were the dominant bacterial genera during the fermentation process. Specifically, Lactobacillus and Lactococcus dominated during the rice-steeping period, with relative abundances of 65.17% and 32.08%, respectively; the subsequent dramatic decreases in these genera were attributable to steaming of the rice. Koji was added to streamed rice after steaming, which caused Leuconostoc, Acinetobacter, Exiguobacterium, Citrobacter, and Pseudomonas to dominate the bacterial community until day 4. The emergence of Staphylococcus, Oceanobacillus, Lactobacillus, Lactococcus, Leuconostoc, Acinetobacter, and Bacillus on the 4th day can be explained by the addition of wheat Qu, with these genera subsequently being present in large proportions throughout the fermentation process. Figure S2 in the supplemental material shows the relative abundances of the phyla and classes of microorganisms during CRW fermentation.

FIG 1.

Heatmap of microbial diversity (A) and relative abundances of dominant microbes (B) during CRW fermentation. (C) Redundancy analysis (RDA) of the fermentation process. Red dots represent the sample times of fermentation, orange arrows point to the different environmental factors, and black arrows represent the dominant microorganisms. Percentages on the axes represent the eigenvalues of principal components. AN, amino nitrogen. (D) Changes in physicochemical properties and BAs during CRW fermentation, with the fermentation process divided into three stages according to the observed variation in these properties.

A temporal profile for the relative abundances of fungal taxa at the genus level indicated that Saccharomyces and Amylomyces were retained at over 90% richness. Saccharomyces comprised the largest proportion of the fungal community throughout the fermentation process, which was more than 72%, followed by Amylomyces, which was attributed to the bilateral fermentation characteristics (fermenting while saccharifying) of CRW. Given the fermentation characteristics of the rice clinker, most communities are likely to originate from fermentation starters and the environment.

Abiotic and biotic factors together shape the taxonomic diversity and functional potential of microbial consortia. Redundancy analysis (RDA) was performed on 20 variables during the fermentation process (15 major microbes and 5 physicochemical properties); these results were significantly different in fermentations of 9 samples (Fig. 1C), and a pattern analysis revealed that the first two variances obviously classified the 9 samples into three groups at a temporal level: canning, main fermentation, and postfermentation. Additionally, the correlation between the dynamics of abiotic variables and microbes was assessed (Fig. 1C). The RDA results show that 7 bacteria genera (Lactococcus, Staphylococcus, Leuconostoc, Oceanobacillus, Lactobacillus, Weissella, and Bacillus) were strongly linked to pH and ethanol; sugar and acid were closely associated with 4 genus taxa (Pseudomonas, Citrobacter, Acinetobacter, and Exiguobacterium). The abiotic variables, including ethanol, pH, and sugar, were almost unchanged after 14 days, with low sugar levels accompanying the low rate of formation of BAs and high levels of amino acids being especially notable during the postfermentation stage (Fig. 1D) (25). Meanwhile, the microbial community structure remained almost invariable with respect to abundance during the postfermentation period (Fig. 1B). This finding suggests that primary succession was driven by changes in resources. More specifically, the dominant resource for supporting microbial metabolism was sugar, and metabolic substrates gradually shifted to refractory substrate consumption over time, which was accompanied by increased energy requirements. Under these conditions, microbes employ a dormancy strategy to decrease metabolic activity and retain viability during the postfermentation stage (26).

We captured 16S rRNA gene sequences to infer the metagenome functional content based on the microbial community profiles using PICRUSt (16). Overall, functional metagenome prediction indicates 295 type functions in Kyoto Encyclopedia of Genes and Genomes (KEGG) level 3, 10 of which are related to BA metabolism (Fig. S3). The weighted nearest sequenced taxon index (NSTI) score of samples was 0.056 ± 0.023 (range, 0.017 to 0.089), as shown in Table S1 in the supplemental material. Furthermore, we selected decarboxylase genes, including agmatine deiminase (aguA), ornithine decarboxylase (odc), lysine decarboxylase (cadA), histidine decarboxylase (hdc), tyrosine decarboxylase (tyrDC), aromatic l-amino acid/l-tryptophan decarboxylase (tdc), and spermidine synthase (speE), and analyzed the contributions of genus taxa, which suggested that 28 bacteria genera contribute significantly (>1%) to the BA metabolic genes during the fermentation process. Pseudomonas was dominant for the aguA gene, but there was a large proportion of unclassified genera that possess aguA; further family-level analysis revealed that Lactobacillaceae and Bacillaceae contributed significantly. Citrobacter contributes a lot to odc and cadA, Acinetobacter is mainly responsible for hdc (91.41%), and tdc genes are primarily from Streptomyces (Fig. 2A). We obtained dominant microbiota (13 bacteria genera; relative abundance >1%), and decarboxylase gene-associated microbiota (28 bacteria genera). Seven genera, including Lactobacillus, Acinetobacter, Bacillus, Exiguobacterium, Citrobacter, Pseudomonas, and Enterobacter, existed in two different microbiota populations (Fig. 2B). Due to their high relative abundances and significant contributions to BA production, they were defined as the main BA-associated microbiota during Chinese rice wine fermentation, and the total relative abundance of these 7 types of bacteria was greater than 30% (Fig. 2C).

FIG 2.

(A) Significant contributions (>1%) of bacterial taxa to the decarboxylase genes. (B) Venn diagram of the dominant BA-related microbiota. Different circles represent different genus categories. (C) Relative abundances of dominant BA-associated bacteria.

Microbial functional analysis with respect to decarboxylase gene expression.

Microbial community taxonomy describes the characteristics of an ecosystem, and differences in BA accumulation rates are attributed to physiological responses or genetic adaptation of the native community under environmental. Consequently, determination of the metabolic mechanism of BAs will be facilitated by understanding the variation in decarboxylase gene expression to link the presence of these genera to function during CRW fermentation. The relationship between bacteria and BAs has been established via the quantitative PCR (qPCR) method used to detect genes involved in amine production (27). Because PUT and TYR accounted for 77.4% of the BA content in CRW, we explored the expression of tyrDC, odc, and aguA genes to elucidate TYR and PUT production during the in situ Chinese rice wine fermentation. The major BA decarboxylase genes were detected (Fig. S4), and the primers used to assess tyrDC, odc, and aguA expression were derived from previous studies (28, 29). During industrial CRW fermentation, PUT is primarily produced via two pathways: (i) the direct decarboxylation of ornithine, and (ii) the aguA pathway, involved in the decarboxylation of arginine into agmatine. The relationship between the level of odc and aguA transcription and the content of substances associated with PUT formation is shown in Fig. 3A. PUT is primarily produced during the early fermentation stage (canning and main fermentation), accumulating to different degrees in 1.5 to 7 days, during which time the odc and aguA pathways work together. Comparing the gene expression profiles, the level of odc expression was higher than that of aguA at the 1.5-, 6-, and 7-day time points, indicating that these three sample points were mainly through the odc pathway. This result was consistent with previous studies on the production of BAs during CRW fermentation (30). Interestingly, the increase in PUT was retarded, even when the levels of odc and aguA transcription sharply increased during the postfermentation stage.

FIG 3.

Relative transcript levels of key genes and the contents of precursor amino acids for putrescine (A) and tyramine (B). (C) Specific formation rates of putrescine and tyramine. Samples were collected on days 1.5, 3, 4, 5, 6, 7, 14, 18, and 21 during fermentation.

Tyramine is directly derived from tyrosine decarboxylation via tyrDC. The tyrDC transcription levels during CRW fermentation are presented in Fig. 3B, together with the variation of TYR and its precursor amino acid. Overall, large fluctuations were observed in the formation rate of TYR. In the main fermentation stage, the expression of the tyrDC gene (Fig. 3B) and the TYR concentration (Fig. 3C) showed an increasing trend. At 1.5 and 6 days, the formation rate of TYR reached the maximum (Fig. 3C), which is mainly ascribed to resource availability (31), and the microbial consortium composition and function related to the BA pathway exhibited a steady pattern. The level of tyrDC transcription at 1.5 days was relatively low, while the specific production rate of TYR was high, indicating a high efficiency of tyrDC. However, in the postfermentation stage, the level of gene transcription sharply increased, which was consistent with the function of decarboxylase in improving metabolic energy and was supported by the observation that carbohydrates were exhausted during the postfermentation stage (Fig. 1D).

Unusually, the high decarboxylase gene expression and precursor amino acid concentration only corresponded to the low rate of PUT and TYR formation during the postfermentation stage, indicating that a combination of stress conditions can activate tyrDC, odc, and aguA gene expression to support cell viability. This result agrees with the previous conclusions made regarding cheese, where decarboxylase gene expression was upregulated under poor conditions and in the presence of precursor amino acids (12, 14). The accumulation of BAs was mediated by primary succession, which suggests the microbial metabolic diversity depends on shifts in resource usage. Glycolysis meets the basic metabolic requirements of cell proliferation that accompany the formation of most BAs in the CRW main fermentation stage. However, when carbon sources were almost depleted, the metabolic perturbations resulting from the altered transcription of genes involved in BA metabolism initiated the transition to dormancy to maintain cell viability during the CRW postfermentation phase. Additionally, the decarboxylation reaction has responded to acid stress at low pH values in the later phase of fermentation. Microbial acid resistance studies have demonstrated that as acidification progresses, ATP is rerouted to generate proton motility, with a decrease in the energy available for biomass synthesis and metabolic flux until microbial proliferation is eventually suppressed (8).

Secondary succession in microbial disturbance during CRW fermentation.

Our survey of the microbial composition during CRW fermentation and the relative transcription levels of key decarboxylase genes suggested that primary succession is driven by changes in resources and that environmental variables are strongly associated with the potential functions of genes. However, a question that remains is how microbial interactions vary with BAs and resources. Measuring microbial interactions directly will always be difficult; thus, constructing a secondary succession (after disturbance) system to assess whether exogenous strains will harm or benefit ecosystem functioning is indispensable. To understand how functional potential varies with taxonomic diversity and microbial interactions in secondary succession, microbial interference experiments were performed in industrial semidry CRW fermentation. Our previous studies have investigated the effect of inoculating single and mixed strains to regulate BAs, suggesting that mixed strains resulted in an approximate 20% reduction from that of individual strains, and BA degradation ability was stronger than the suppression effect of low decarboxylase activity (25). Subsequently, we conducted mixed-strain interference at different ratios within industrial CRW fermentation, which resulted in a marked decrease in BA concentration (Fig. 4A). Compared with the control group (uninoculated), the total BA content was reduced by 31.58, 46.49, and 31.54% at coinoculation ratios of 1:1, 1:2, and 1:3, respectively. Thus, the strength of the disturbance had an effect on the microbial function.

FIG 4.

(A) The total BA, putrescine, and tyramine concentrations during microbial interference experiments. L. plantarum (ACBC271) and S. xylosus (CGMCC1.8382) were coinoculated at ratios of 1:1, 1:2, and 1:3, respectively. (B) Changes in sugar content in the uninoculated and coinoculated groups during CRW fermentation. (C to F) The relative abundances of bacteria in the uninoculated group (C) and the groups coinoculated with L. plantarum (ACBC271) and S. xylosus (CGMCC1.8382) at ratios of 1:1 (D), 1:2 (E), and 1:3 (F), respectively.

Microbial interactions and environmental factors determine the function of the community. The microbial diversity of the CRW fermentation process under disturbance was evaluated to understand the drivers of secondary succession. Consistent with primary succession, the depletion of carbon sources divided microbial diversity succession into three phases (Fig. 4B). In the postfermentation stage, the relative abundance of the community was essentially unchanged and sustained at a high level, and the concentration of BAs was largely unchanged. Compared with the control group, the coinoculation group had higher residual sugar levels and a lower ethanol content, which could be due to competition between the exogenous strains and Saccharomyces weakening the ability of the latter microbe to convert sugar into ethanol (Fig. S6). Specifically, the relative abundance of Saccharomyces did not change after the 7th day in the coinoculation (1:2) group (Fig. S6C), which may be attributed to the earlier microbial dormancy. Concretely, four groups showed diverse relative abundances of microorganisms according to the divergent proportion of supplementary microbes, and the relative abundances of Lactobacillus and Staphylococcus held dominance in the coinoculated group throughout the fermentation process. In the group coinoculated with Lactobacillus and Staphylococcus at a ratio of 1:1, the relative abundance of Lactobacillus sharply increased, reaching 50.65% on the 6th day (Fig. 4D). An explanation for this result could be that Lactobacillus possesses a strong inhibitory effect toward other genera compared to Staphylococcus. This finding may imply that Lactobacillus occupies the most extensive ecological niche in the CRW fermentation system. Furthermore, the relative abundances of Staphylococcus in the 1:2 and 1:3 coinoculated groups were almost equal, and the richness of other genera in the 1:3 group, except Lactobacillus, did not notably decrease and was greater than that of the 1:2 group (Fig. 4E and F). This result indicates that the microbial ecosystem in the CRW fermentation is likely to reach equilibrium in the 1:2 coinoculated group. Alternatively, because the resources available to Staphylococcus in the CRW fermentation system may be limited, we assume that the saturation of niche resources can only support a certain amount of Staphylococcus. The bacterial genera Citrobacter, Acinetobacter, and Exiguobacterium, whose abundance and contribution to decarboxylase were significant, had the lowest relative abundance in the 1:2-coinoculated group.

The transaminase and polyketone oxidase carried by Staphylococcus xylosus and the low decarboxylase activity of Lactobacillus plantarum result in a remarkable reduction in BAs during mixed fermentation. Microbial interactions directly affect metabolism after disturbance; Spearman’s rank correlation analysis was performed to identify the beneficial or antagonistic relationships between various microflora at the genus level (Fig. 5). A positive number is indicative of a beneficial relationship, and a negative number represents an antagonistic relationship. Overall, the inhibitory effect of Lactobacillus on other genera in the coinoculated fermentation group was notably stronger than that observed in the uninoculated group. In the 1:2 coinoculation group, BA-associated bacteria, including Exiguobacterium and Pseudomonas, inhibited the strongest suppression by Lactobacillus (ρ = −0.867 and ρ = −0.782, respectively; P < 0.05) (Fig. 5A). These negative interactions between species may be ascribed to the competition for adjacent niches as well as molecule-based signaling between lactic acid bacteria and other genera that inhibit growth.

FIG 5.

Spearman rank correlations showing the beneficial and antagonistic relationships between genera and Lactobacillus (A) as well as Staphylococcus (B). A positive number is indicative of a beneficial relationship, and a negative number indicates an antagonistic relationship; significant positive and negative associations are indicated with a bold boundary. Only the first 13 bacterial genera were analyzed.

Furthermore, the positive effect of Staphylococcus toward Lactobacillus, Acinetobacter, and Citrobacter in the uninoculated group was transferred into the suppression effect. Staphylococcus in the 1:2 mixed fermentation group showed the strongest inhibitory effect toward Lactobacillus (ρ = −0.115) and Citrobacter (ρ = −0.188), as well as the lowest positive effect on Enterobacter and Bacillus (Fig. 5B). Therefore, the high inhibitory effect and low promoting effect of exogenously added strains on specific bacteria provided evidence of the obtained BA formation contributors. Secondary succession after disturbance is driven not only by changes in resource availability but also by persistent microbial interactions, which create a new equilibrium in the disturbed state and alter microbial traits and functions. Thus, the regulation of BAs by the exogenous addition of functional microorganisms is feasible because even if microbial composition is sensitive to disturbances, the community may still be resilient and can quickly rebalance to a new steady state to achieve a lower concentration of BAs and retain the quality of the product.

DISCUSSION

Disentangling the relationships between microbial community assembly and potential function is a major challenge within microbial ecology. In this study, industrial CRW fermentation and microbial disturbance experiments were performed that accounted for BA accumulation kinetics coupled with abiotic conditions, transcript analysis, and population size to profile the primary and secondary microbial successions that mediate BA formation. In addition, the patterns of resource drivers and microbial interactions, as well as the equilibrium mechanism of microbial systems, were investigated. Succession leading to stability is a primary and common principle in community ecology (32). A stable community has a high degree of self-equilibrium, being able to maintain the stability of the community and resist external influences that disrupt the stable state through the use of regulatory mechanisms based on the adaptation of the community members (33). When major changes occur, such as microbial disturbances, they may disrupt homeostasis and lead to secondary succession, after which the resilience of microbial consortia can regulate the metabolism to bring about a new balance for a steady system.

Functional microbiota cannot often be compatible with the dominant community due to the strain dependence of metabolites (34). That is, although a genus may be less abundant, it may correspond strongly with functional genes; for instance, Pediococcus contributed significantly to aguA (2.1%) but had low abundance (<1%), as shown in Fig. 2. Together, relative abundance and functional contributions signify the characterization of key microbiota, which can promote further understanding of the mechanisms that affect the assembly and dynamics of ecosystems. We provide an approach for identifying corelated functional microbiota based on the intersection of genomic function prediction and relative abundance. After we analyzed the microbial composition, the functional features of the microbiota associated with BA formation were addressed using PICRUSt. Notably, since fractional operational taxonomic unit (OTU) annotation may not be matched in the Greengenes database, the PICRUSt results are not comprehensive enough to replace meta-transcriptomics (16); thus, the molecular investigation needs to determine the changes in decarboxylase gene expression or mRNA transcript levels for BA-associated functional genes.

Bacterial and fungal colonization and acclimation are examples of the important influence of the microbial community on material circulation. Microorganisms from the environment and fermentation agents represent complex colonization pools of yeasts, molds, and bacteria that attach to the CRW fermentation system and eventually form a dense multispecies community. Microbial community divergence was delineated into different stages that are closely associated with sugar consumption (Fig. 1B to D), suggesting that primary succession is driven by resource availability. The decrease of BA formation ratio and high relative abundance of BA-related microbes can provide evidence of microorganisms entering dormancy during the CRW postfermentation phase. Microbes have evolved a variety of genetic and cellular mechanisms involved in dormancy that may help explain the ecological and evolutionary phenomena where microbes decrease their metabolic activity and maintain viability with respect to succession dynamics (26). Compared with the uninoculated group, the residual sugar content was higher in the coinoculated fermentation group (Fig. 4B), indicating that secondary succession had an effect on the resource utilization and that disturbances provoked an earlier dormancy. In microbial consortia, local interactions among the members of a community occur that play crucial roles in shaping the functional potential of the community (Fig. 5).

The high content of amino acids, the high abundance of microbes, and the observed increase in gene transcription corresponded with a low BA formation rate during the postfermentation stage, during which time there are low levels of carbon sources (Fig. 1A and D; Fig. 3) (25). An explanation for this result could be that odc and aguA were involved in energy metabolism to support basic growth rather than BA production. The aguA pathway has been shown to facilitate the growth of lactococci under conditions of nutrient depletion (35). In addition to the adaptability to a niche requiring the ability to resist the nutrition-poor environment and activate the bacterial metabolic pathways to supply energy, we speculate that the decarboxylase activity might not be sufficient to fulfill the requirements of the BA accumulation pathway. The genetic clusters encoding decarboxylases involved in the BA biosynthesis pathway are complicated. For instance, tyrosine decarboxylase loci typically contain genes encoding tyrosine decarboxylase, tyrosyl tRNA synthetase, tyrosine/tyramine permease, and Na⁺/H⁺ reverse transporters (36). The environmental pressure that occurs during the CRW postfermentation phase may retard or limit the activities of permeases and transporters.

Microbial interactions include competition and mutual benefit, and the utilization of bacteria with active or passive migration may affect their ecological succession (37). Over the course of microbial succession, the ability of microorganisms to survive and reproduce depends, in part, on their functional characteristics that confer colonization potential, competitive advantage, dormancy, or stress tolerance (38). However, how these processes combine to generate specific outcomes in microbial communities, especially under circumstances of multiple natural microbial combinations, is still poorly understood. Thus, relevant new studies need to be designed to establish a broader model to explore the relationship between niche distribution and microbial evolution in traditional fermentation systems and to significantly advance fundamental information about microbial interactions within and beyond ecological systems.

MATERIALS AND METHODS

Microbial interference experiment in CRW fermentation.

The industrial fermentation of CRW was performed at the Baipu Chinese Rice Wine Co., Ltd. (Nantong, China), and the traditional semidry CRW production process is shown in Fig. S1 in the supplemental material. This study used L. plantarum (strain ACBC271) and S. xylosus (strain CGMCC1.8382) for microbial interference experiments. L. plantarum (ACBC271) and S. xylosus (CGMCC1.8382) were coinoculated on the 3rd day of the experiment at ratios of 1:1, 1:2, and 1:3. Strains were mixed in approximately equal concentrations of 3.5 × 10⁷ CFU ml⁻¹ and inoculated into a CRW fermentation system. Twenty-milliliter aliquots of each sample from 3 batches at 9 time points (1.5th, 3rd, 4th, 5th, 6th, 7th, 14th, 18th, and 21st days) were collected and stored at 4°C after filtering for subsequent analysis.

BA concentration was quantified by reverse-phase high-performance liquid chromatography (HPLC) using an instrument equipped with a diode array spectrophotometer (Hitachi, Japan) and derivatization with dansyl chloride according to the method previously reported (30). The physical and chemical indexes (total acid, sugar, ethanol, and amino nitrogen) during the fermentation process of CRW were determined according to the national standards of China.

DNA extraction, amplification, and sequencing.

To track the relative abundance within in situ and the coinoculation CRW fermentation, 4 ml of samples was centrifuged (10,000 × g for 3 min at room temperature), and the supernatants were discarded. Microbial genomic DNA was extracted using an E.Z.N.A. soil DNA kit (catalog no. D5625; Omega Bio-Tek, USA) following the manufacturer’s instructions. The V3-V4 hypervariable region of the 16S rRNA gene was amplified using the universal primers F341 and R805 to prepare bacterial amplicon libraries. The same approach applied V43F/EukV4R to amplify the 18S rRNA gene of fungi (39). The amplicon data were analyzed using QIIME (version 1.8.0) (40). Briefly, the raw sequences present at quality scores <30 were removed and sequences >200 bp were used for subsequent analysis. Those sequences that incompletely matched the PCR primer, had unassigned tags, or had an N base were eliminated. The UCHIME algorithm was utilized to eliminate chimeric sequences. Reads were merged and quality filtered with custom scripts and were then clustered into diverse operational taxonomic units (OTUs), defined as having a ≥97% 16S rRNA gene sequence similarity, using UCLUST.

qPCR analysis of decarboxylase gene expression.

The two-step reverse transcription-PCR (RT-PCR) was designed for the relative quantification of key decarboxylase gene transcripts. Specifically, RNA was extracted from each sample in Chinese rice wine fermentation, and to generate cDNA by reverse transcription, qPCR was performed with 16S rRNA as an internal reference gene by using IQ SYBR green supermix (Bio-Rad, Marnes-la-Coquette, France). Primers were designed in conserved regions of the tyrDC, aguA, and odc genes using Oligo 6.0 (Molecular Biology Insights, Inc., USA), and primer specificity was confirmed by using BLAST at NCBI. Four pairs of qPCR primers were used in this study (Table 1): 16SF and 16SR were used as reference gene primers, aguAF-aguAR and tyrDCF-tyrDCR were described in previous studies (28, 29), and primers odc1F and odc1R were obtained according to the sequence alignment of Oenococcus oeni (BR14/97, SA. B 034, DIV 5.74, DBA, and DIV 7A.3) and Staphylococcus lugdunensis (HKU10-04, HKU10-07, HKU10-01, and HKU10-02). Subsequently, reactions were performed by using 10 μl SYBR green qPCR master mix, 0.4 μl of each primer, 2 μl of the cDNA preparation, and 7.2 μl double-distilled water (ddH₂O). The PCR program was comprised of initial denaturation at 95°C for 5 min and 40 cycles at 95°C for 30 s, 55°C for 30 s, and 72°C for 30 s, followed by a melting curve analysis. For each assay, the calculation of the decarboxylase genes expression from plotting the threshold cycle (C_T) values was performed by using the calibration curve. The slopes of the curves and the regression coefficients (R²) are reflected in the consistency of the amplification rates of the target gene and the control, as shown in Table 2 and Fig. S5.

TABLE 1.

PCR primers used in this study

Primer pair	5′→3′ sequence	Amplicon size (bp)	Source or reference no.
odc2F and odc2R	CTACGGATTTATTGGTGG/GTGGTTGATGTAGCGTTT	322	This article
16SF and 16SR	CCTACGGGAGGCAGCAG/ATTACCGCGGCTGCTGG	196	This article
agdiF and agdiR	ATGCCCGGTGAATTTGAA/TTGCGCTGGTTTAGCACC	90	24
odc1F and odc1R	TGTAATGCCGATGTTGATTTAG/CGAAATATGTTTTGTCAGCATT	106	This article
tyrDCF and tyrDCR	TGAGAAGGGTGCCGATATTC/GCACCTTCCAACTTCCCATA	141	25

TABLE 2.

Standard curves of relevant qPCR genes

Gene detected	Equation	R2	Amplification rate (%)
16S	y = −3.363x + 19.998	0.995	98.322
agdi	y = −3.303x + 24.703	0.999	100.783
odc1	y = −3.394x + 25.574	0.999	97.074
tyrDC	y = −3.341x + 23.984	0.998	99.218

Functional metagenome predictions.

To better understand the potential functional contributions of the observed shifts in microbial diversity, functional profiles of the bacterial communities were predicted using the PICRUSt algorithm. Inferred metagenomics and predicted functional analysis were initiated by generating closed reference sequence clustering OTUs in QIIME. The quality-filtered, demultiplexed sequences were clustered by 97% similarity. Representative sequences were picked and aligned with the sequences from the Greengenes reference collection (gg_otus_13_5.tar.gz) to assign taxonomy (41). The resulting OTU table was used to generate inferred metagenome data using the online Galaxy interface for PICRUSt with default settings. Briefly, the abundance values of each OTU were normalized to their respective predicted 16S rRNA copy numbers and then multiplied by the respective predicted gene counts for metagenome prediction. The resulting core output was a list of the Kyoto Encyclopedia of Genes and Genomes (KEGG) orthologues and predicted gene count data for each sample. The nearest sequenced taxon index (NSTI) score, which is the sum of phylogenetic distances for each OTU between its nearest relative with a sequenced reference genome, measured in terms of substitutions per site in the 16S rRNA gene and weighed according to the frequency of that OTU, was used as an indicator for the accuracy of PICRUSt.

Statistical analysis.

Each experiment was conducted for three batches, and relevant data in figures and text are presented as the means with standard deviations (mean ± SD) for samples. Statistical differences of the total acid, sugar, ethanol, amino nitrogen, and BAs among noninoculation and coinoculation samples were obtained by the application of one-way analysis of variance (ANOVA). To analyze the relationships among microbial communities, Spearman’s correlation between the exogenous microbes and the abundant genera was calculated using SPSS 19.0 (SPSS Inc., USA). Redundancy analysis (RDA) of dominant microbes and physicochemical parameters during the fermentation process was performed via CANOCO 5.0 (Microcomputer Power, Ithaca, NY, USA). Figures were generated by using Origin 8.0 (OriginLab, MA, USA).