Dynamic succession of the fermentation microbial community and its contribution to aroma in pomelo wine

Abstract

The unique flavor quality of pomelo wine is closely related to its complex fermentation process and metabolic activities. The aim of this study was to gain a deeper understanding of the microbial community succession in pomelo wine fermentation and its specific contribution to aroma formation. This was achieved by analyzing the dynamic succession of microbial communities during the fermentation process of pomelo wine and then evaluating the role of hub taxa microbiota in the generation of key aroma compounds in the wine. Bacillus, Pichia, Komagataeibacter, Pediococcus, Zasmidium, Penicillium, Saccharomyces, Weissella, and Gluconobacter were identified as hub taxa microbiota, and their relative abundances showed a correlation with multiple key volatile aroma metabolites detected during pomelo wine fermentation. These findings of functional zoning and aroma contribution evaluation of the pomelo wine fermentation microbial community provide a scientific basis for realizing the precise regulation of the fermentation process, standardization and optimization of pomelo wine flavor quality, and thus enhancing product competitiveness.

Highlights

• •Exploring the relationship between microorganisms and fermentation flavor of pomelo wine.
• •It provides a scientific basis for broadening the horizontal processing of pomelos.

1. Introduction

Pomelo (Citrus grandis L.), commonly known as Chinese grapefruit or chakotra in Hindi, is a citrus fruit belonging to the Rutaceae family. Pomelo is widely distributed in areas with a mid-subtropical monsoon climate. It has a cultivation history of more than 3000 years in China, with Jiangxi, Fujian, Guangdong, and Guangxi as the main cultivation regions. The fruit consists of a yellow or green peel and a white or pink pulp. The taste is characterized as sweet and slightly sour. In 2020, the global pomelo production was 9.34 million tons, 53.9% of which was produced in China (Shangguan et al., 2023).

Jiangxi Jinggang honey pomelo represents a specific category of high-quality pomelo varieties cultivated in Ji'an City, Jiangxi Province, which have a honey-like flavor. There are three characteristic varieties of Jinggang honey pomelo: Taoxi honey pomelo, Jinsha pomelo, and Jinlan pomelo. Owing to its plump pulp, sweet-and-sour and juicy taste, and slightly bitter aftertaste, Jinggang honey pomelo is favored by consumers. These pomelo varieties are also used to prepare by-products such as jam, fruit juice, fruit wine, and candied fruit, all with a high market demand. In addition, compared with other pomelo varieties, Jinggang honey pomelo is rich in various health-promoting substances such as vitamin C, flavonoids, limonoic acid, and carotenoids. These components have powerful antioxidant effects, can neutralize free radicals, and protect cells from oxidative damage, ultimately enhancing immunity to facilitate infection resistance (Qin et al., 2018). According to “Compendium of Materia Medica,” pomelos are effective in resolving phlegm and relieving cough, along with effectiveness in treating loss of appetite and indigestion. Moreover, long-term pomelo consumption has been shown to improve skin condition. The pulp and peel of pomelos contain bioactive substances, namely bio-glycosides, which can reduce blood viscosity and decrease the formation of blood clots. Pomelos have also been reported to have a preventive effect on cerebrovascular diseases such as cerebral thrombosis and stroke (Tan et al., 2011).

In addition to fresh consumption, the processing by-products of pomelos, especially the peel and pulp, have untapped potential in value-added applications, including alcoholic fermentation. Pomelo wine is an emerging functional beverage that combines the natural antioxidants of the fruit with the bioactive metabolites produced during the fermentation process, offering a new strategy for waste valorization and sustainable agriculture. Therefore, using pomelo juice for fermentation and wine-making not only retains the abundant nutrients of the fruit but also enriches the flavor and quality of the obtained wine.

The flavor and aroma characteristics of fruit wine are key factors influencing consumer preferences. Pomelo wine is highly favored by consumers owing to its unique citrus aroma and flavor. These complex flavor compounds mainly come from the fruit components, metabolic activities of microorganisms during fermentation, and the subsequent aging process (Wu et al., 2025). In particular, the microorganisms, namely yeast and bacteria, in the fermentation process play a core role in the formation of flavor compounds in fruit wine (James et al., 2023).

However, the specific functions of different microorganisms in fruit wine fermentation are difficult to elucidate owing to the complex and dynamic process involving interactions and niche competition among multiple microbial species (Wu et al., 2025). For example, during the fermentation process of wine, the microbial community on the surface of grapes will transfer to the fermentation broth, thereby having a strong influence on the final quality parameters of the wine (Chen et al., 2023). The fermentation process of grapefruit wine also involves the succession of microbial communities, which convert sugar into alcohol through complex biochemical reactions, producing various volatile aroma compounds in the process. Although the importance of microorganisms in the formation of fruit wine aroma has been widely recognized, the so-called “functional zoning” within microbial communities and their contribution mechanisms to specific aroma compounds remain unclear (Jiang et al., 2025).

Functional zoning refers to the occupation of specific ecological niches by different microbial groups or species in a microbial community, along with the execution of specific metabolic functions, thereby affecting the types and contents of aroma substances. Therefore, analyzing the functional zoning of various microbial groups in complex fermentation microbial communities can provide greater insight into their specific contributions to aroma compounds in the final product (Wei et al., 2022). In turn, understanding these ecological mechanisms is crucial for optimizing the fermentation process of grapefruit wine, regulating its flavor characteristics, and developing new grapefruit wine products with specific sensory attributes (Wang et al., 2024).

Therefore, the aim of this study was to explore the functional zoning characteristics of the fermentation microbial community in pomelo wine by combining high-throughput sequencing, metabolomics analysis, and bioinformatics methods. We further evaluated the contributions of different microbial groups to key aroma compounds in the wine. These results will not only provide scientific guidance for the brewing process of pomelo wine, but can also help to expand our general understanding of the complex interactions between the microbial ecology and flavor chemistry of fruit wine.

2. Material and methods

2.1. Materials and reagents

Hydrochloric acid (SCR, Tianjin, China) was diluted with deionized water at a 1:1 volume ratio. Sodium hydroxide (NaOH; ≥98%, SCR) was prepared as a 200 g/L solution. The glucose standard solution (2.5 g/L) was prepared from glucose (≥99.5%, SCR). Methylene blue (C₁₆H₁₈ClN₃S; ≥98%, Aladdin, Shanghai, China) was prepared as a 10 g/L solution. Copper sulfate (CuSO₄; ≥99.5%, SCR) was prepared as a 0.05 g/mL solution. Sodium hydroxide solution (0.05 mol/L) was prepared from NaOH (≥98%, SCR). Sodium chloride (NaCl; ≥99.5%, SCR) and Seignette salt (potassium sodium tartrate; ≥98%, SCR) were used as received. Pectinase was obtained from Zhongchen Biotechnology Co., Ltd. (He’nan, China). Citric acid/sodium citrate was obtained from Yingxuan Industrial Co., Ltd. (Shandong, China). The mixed fermentation koji was obtained from YADA (Hunan, China).

2.2. Sample preparation and fermentation

In October 2024, mature and healthy Jinggang honey pomelo fruits were selected from the production area in Ji'an City, Jiangxi Province and brought back to the laboratory for fermentation experiments. The fermentation process was carried out at 20 °C. The pomelo was peeled and the juice was extracted from the flesh with a juicer. The juice was then added to a 30-L fermentation tank with pectinase (1 g/L), the soluble solids content was adjusted to 22°Brix, and the pH was adjusted to 4.0. Subsequently, according to the instructions from the mixed fermentation koji product, 5.2 g/L of the yeast was placed in a small beaker containing a small amount of pure water. The beaker was placed in a 30 °C water bath for activation, transferred to a fermentation tank and mixed thoroughly. The static fermentation process was concluded after 14 days. The estimated alcohol content at the end of fermentation was 10 v/v% (representing a semi-sweet wine type). Samples were collected separately in batches on days 0, 1, 4, 7, 10, and 14 of fermentation (n = 18) and frozen at −80 °C for subsequent analysis.

2.3. Physiochemical analysis

Multiple physical and chemical parameters of the samples were measured, including pH value, soluble solids (Brix), total acidity, total sugar content, and ethanol concentration. A pH meter was used to measure the pH value of the solution. The total acidity (TA) was determined by potentiometric titration, with pH 8.2 as the titration endpoint. The total soluble solids (TSS) content was detected using a hand-held refractometer. The ethanol concentration was measured using a CJM-091 alcohol meter (China). The total sugar content (TS) was measured by the direct titration method. Each measurement data has three biological replicates.

2.4. Sample processing, DNA extraction, and taxonomic identification

A sample (0.2–0.5 g) from each fermentation time was added to extraction lysis buffer and ground at 60 Hz. DNA was extracted from the ground sample using the MagBeads FastDNA Soil Kit (116564384) and quantified with a Nanodrop spectrophotometer.

For bacteria, the highly variable V3–V4 region of the 16S rRNA gene was selected for sequencing. For fungi, the ITS1 region was selected for sequencing. The V3–V4 regions of the 16S rRNA genes were amplified by polymerase chain reaction (PCR) using the primers 338F (5′-ACTCCTACGGGAGGCAGCA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′), whereas the ITS1 region was amplified with the primers ITS5-1737F (5′-GGAAGTAAAAGTCGTAACAAGG-3′) and ITS2-2043R (5′-GCTGCGTTCTTCATCGATGC-3′) (Li et al., 2023). NEB Q5 DNA high-fidelity polymerase was used in all PCR runs. The products were subject to 2% agarose gel electrophoresis for visualization. The target fragments were excised and recovered using the Axygen gel recovery kit.

2.5. Sequence analysis

The microbial taxa were identified using QIIME2 version 2024.5 according to the process that was modified and improved by Bolyen et al. (2019). In brief, the raw sequence data were decoded using the demux plugin, and the primers were removed using the cutadapt plugin. The sequences were quality-filtered, denoised, and merged, and chimeras were removed using the DADA2 plugin. Non-singleton amplicon sequence variants (ASVs) were aligned using mafft, which were used to construct a phylogenic tree with the fasttree2 tool. Taxonomic assignment was performed using the classify-sklearn naïve Bayes taxonomy classifier within the QIIME2 feature-classifier plugin. Bacterial ASVs were classified against the SILVA 16S rRNA gene database (Release 140), whereas fungal ASVs were classified against the UNITE ITS database (Release 9.0).

2.6. Analysis of volatile compounds

The volatile compounds in the samples were analyzed by headspace solid-phase microextraction coupled with gas chromatography–mass spectrometry (HS-SPME-GC–MS). The conditions for GC analysis were as follows: helium was used as the carrier gas with a constant flow rate of 1.5 mL/min passed through the column. The initial temperature was maintained at 40 °C for 3.5 min, and then the temperature was programmed to increase to 100 °C at a rate of 10 °C/min, followed by an increase to 180 °C at a rate of 7 °C/min, and finally an increase to 280 °C at a rate of 25 °C/min; the system was held at this temperature for 5 min. The inlet, transfer line, ion source and quadrupole temperatures were set to 250 °C, 250 °C, 230 °C, and 150 °C, respectively. Electron bombardment mode was used with an energy of −70 eV. Mass spectral data were acquired in selective ion monitoring mode.

Quantitative analysis:The quantitative method is the internal standard semi quantitative method, which refers to a suitable compound standard added to the sample during the quantitative analysis process (a. The original sample does not contain any components; b. The retention time of the test substance should be close to, but not overlapping with; For high-purity standard substances or substances with known content; D has a certain chemical stability under given chromatographic conditions, and its measured value is the reference basis for calculating the content of the tested component. In quantitative analysis, in order to ensure that the operation meets the requirements of the detection method, an appropriate compound standard is often selected as the accompanying reference material, and a known amount is added to the sample while detecting both the tested component and the reference material. According to the analysis of literature research results, isotopes are currently one of the commonly used internal standard reagents. Based on this, we choose an isotopic internal standard and calculate the relative content of VOCs in the sample according to the following formula based on the reference calculation method.

• (1)
  Formula for calculating the relative content of compounds in solid

  • •
    $$X_i=\frac{V_s\times C_s}{M}\times \frac{I_i}{I_s}\times 10^{-3}$$

X_i is the content of compound i in the test sample (μg/g); V_s is the volume of the internal standard added (μL); C_s is the concentration of the internal standard substance (μg/mL); M is the amount of the sample to be tested (g); $I_s$ is the peak area of the internal standard substance; I_i is the peak area of compound i in the sample to be tested.

• (2)
  Formula for calculating the relative content of compounds in liquid samples

  • •
    $$X_i=\frac{V_s\times C_s}{V}\times \frac{I_i}{I_s}\times 10^{-3}$$

X_i is the content of compound i in the test sample (μg/mL); V_s is the volume of internal standard substance added (μL); C_s is the concentration of the internal standard substance (μg/mL); V is the volume of the sample to be tested (mL); $I_s$ is the peak area of the internal standard substance; I_i is the peak area of compound ì in the sample to be tested.

Construction and analysis of self built database:

• (1)
  Qualitative analysis of database construction process: Through the Agilent Qualitative System MassHunter software, the deconvolution parameters were set with a peak width of 20, and the requirements for resolution, sensitivity, and chromatographic peak shape were all set to medium. The minimum value of the matching factor was set to 70. Based on the obtained mass spectral data, compound identification was performed by comparing with the mass spectra of standard substances provided by the NIST (2020) spectral library. Subsequently, the retention indices (RI) of each volatile compound were calculated. Using an n-alkane mixture (C7–C40) as the standard, gas chromatography–mass spectrometry (GC–MS) analysis was conducted under the same chromatographic conditions, and the retention indices of the compounds were calculated according to the following formula for comparison with the RI values reported in the literature. In 1963, Van Den Dool et al. introduced the concept of linear programmed temperature retention index through calculation (the detailed derivation process is omitted).

  • •
    $$\mathrm{RI}=10\text{0Z}+100\left[\mathrm{TR}\left(\mathrm{x}\right)-\mathrm{TR}\left(\mathrm{z}\right)\right]/\left[\mathrm{TR}\left(\mathrm{z}+1\right)-\mathrm{TR}\left(\mathrm{z}\right)\right]$$

In the formula: TR(x), TR(z), and TR(z + 1) represent the retention temperatures of component x, n – alkane with carbon number z, and n - alkane with carbon number z + 1, respectively. And TR(z) < TR(x) < TR(z + 1).

At the same time, further qualitative confirmation is carried out by combining with literature reports (Yuan et al., 2024), and a specific volatile broad - target database is established through the above methods, which is consistent with the above requirements.

• (2)Qualitative analysis in the detection process: The SIM (Selected Ion Monitoring) detection method is adopted. All ions to be detected in each group are detected separately in time segments according to their elution order. If the retention time of the detected chromatographic peak is consistent with the standard reference, and the selected ions all appear in the sample mass spectrum after background subtraction, the substance is determined to be identified (Yuan et al., 2024). The quantitative ions are selected for the integration and correction of the chromatographic peak to enhance the accuracy of quantification.

2.7. Calculation of the relative odor activity value (rOAV)

Relative odor activity value (rOAV) is a method established by combining the sensory threshold of compounds to determine key flavor compounds in food. It is used to elucidate the contribution of each aroma compound to the overall aroma characteristics of the sample. In recent years, rOAV has been increasingly applied by scholars to determine key flavor compounds in various types of food. Generally, rOAV >1 indicates that the compound makes a direct contribution to the flavor of the sample. The calculation formula is as follows:

• •
  $${\text{rOAV}}_{\mathrm{i}}=\frac{{\mathrm{C}}_{\mathrm{i}}}{{\mathrm{T}}_{\mathrm{i}.}}$$

In the formula, rOAV_i represents the relative odor activity value of compound i, and C_i represents the relative content of the compound (μg/g or μg/mL); Ti Threshold for compounds (μg/g or μg/mL).The higher the rOAV, the greater the contribution of the component to the overall flavor characteristics of the sample. In the analyzed samples, components with an rOAV value of no less than 1 were regarded as key flavor compounds, while components with an rOAV value between 0.1 and 1 were considered to play a crucial regulatory role in shaping the overall flavor characteristics of the sample. The main source of threshold data is based on a large number of references, and by summarizing the threshold values and frequency of occurrence of compounds in literature, books, and domestic and foreign databases (one compound may correspond to multiple thresholds), the threshold with the highest frequency of occurrence and closer age was selected.

2.8. Statistical analysis

All statistical processing was performed using SPSS 21.0 software. The diversity of the microbial community was assessed based on the Chao 1 index and Shannon indices; the relevant functions in the R language package were combined for calculation to reveal the richness and evenness characteristics of the microbial community. Principal coordinate analysis (PCoA) was employed to analyze the overall structural differences and distribution trends of the microbial community, in which the arrangement of points in the multi-dimensional space was used to reflect the similarities and differences among different communities. Origin 2021pro software was used to draw the distribution histograms of microbial communities and volatile components. The correlation between brewing process parameters and hub taxa microbiota communities was explored using redundancy analysis (RDA) with the “vegan” R package, In addition, Gephi (v0.9.5) software was used to construct a network topology map to reveal the interaction relationships between the hub taxa microbiota communities and characteristic aroma substances during the pomelo wine fermentation process.

2.9. Correlation network analysis

Correlation network analysis is a powerful tool in microbiology that is commonly used to reveal the interactions and ecological relationships between species within microbial communities. By constructing a correlation network, it is possible to identify co-occurring or mutually exclusive microbial communities, thereby inferring potential ecological functions and interaction mechanisms. The threshold selection in correlation networks has a decisive impact on the topology and biological interpretation of the network. R > 0.8 was set as the correlation threshold in this study, which is considered a strict threshold that is used to identify the strongest microbial interactions. |R| ≥ 0.6 reflects positive or negative correlations with moderate strength, which can still comprehensively capture the interactions between microorganisms.

3. Results and discussion

3.1. Physical and chemical parameters in wine brewing

Table 1 summarizes the changes in pH, TA, TS, TSS, and alcohol content during pomelo juice fermentation. With the extensive reproduction of microbial populations, enzymatic reactions mediated by microbial enzymes act on carbohydrates, proteins, and lipids, promoting the transformation of these macromolecular organic substances (Carpena et al., 2020). During the fermentation process, catabolites and polymer metabolites will appear (Yuan et al., 2022). Therefore, these metabolites have an impact on the surrounding brewing parameters. At fermentation termination, the pH, TS, and TSS decreased, whereas TA increased. The alcohol content surged significantly to 8.963% (v/v) during days 0–7 of fermentation. Previous studies indicated that this rapid ethanol accumulation correlates with microbial exponential growth phase, where intense metabolic activity drives accelerated sugar-to-ethanol conversion (He et al., 2021). This pattern aligns with findings in pomelo wine fermentation, where the ethanol concentration exhibited a steep rise within the first 4 days of primary fermentation (Zhao, Zhu, & Dong, 2019). However, the TS content did not show a monotonic trend in fermentation, which is likely the result of substrate complexity, microbial community succession, and the combined action of enzymatic/non enzymatic reactions. During the initial stage (from day 0 to day 4), microorganisms may consume sugars through metabolic pathways, resulting in the observed decrease in TS from 115.2 g/L to 92.75 g/L. Subsequently there was an increase in the TS content, which may be related to the complex fermentation system. If the raw materials contain a certain amount of pectin, starch, or other complex carbohydrates, as the fermentation process progresses, some enzyme-producing microorganisms may secrete extracellular hydrolytic enzymes, gradually reducing large molecular polysaccharides into soluble monosaccharides or disaccharides, resulting in a periodic increase in the measured “total sugar” content. Previous studies have shown that in mixed fermentation systems, there is metabolic division of labor and material exchange between dominant and auxiliary strains, which may lead to the transient accumulation of oligosaccharides (Lee et al., 2024; Xie et al., 2024).

Table 1.

Physicochemical indicators during the fermentation process of pomelo wine.

Physicochemical indices	Fermentation process
	0d	1d	4d	7d	10d	14d
pH	4.200 ± 0.016a	3.960 ± 0.024ab	3.940 ± 0.008b	3.843 ± 0.005c	3.837 ± 0.005c	3.920 ± 0.008b
TA (g/L)	4.670 ± 0.141d	5.363 ± 0.009c	7.270 ± 0.141b	7.603 ± 0.189ab	7.770 ± 0.245a	7.870 ± 0.308a
TS (g/L)	115.197 ± 1.952c	115.197 ± 1.952a	92.750 ± 1.131c	102.867 ± 1.226b	116.833 ± 2.245a	87.800 ± 2.121d
TSS (%)	19.167 ± 0.236a	15.933 ± 0.330b	8.600 ± 0.22c	8.400 ± 0.082c	7.167 ± 0.236d	7.100 ± 0.082d
Alcohol (v/v,%)	n.d.	2.25 ± 0.078d	6.477 ± 0.060c	8.963 ± 0.175c	10.543 ± 0.066b	10.723 ± 0.033a

Note: Data are presented as mean ± standard deviation (n = 3); a-d: Different lowercase letters indicate significant differences between the samples (P < 0.05, one- way ANOVA); Total acid is expressed as tartaric acid; n.d.: not detected.

3.2. Dynamic changes and composition of microorganisms during the fermentation process of pomelo wine

3.2.1. Changes in microbial community structure during fermentation

PCA was used to compare the differences in microbial community structure and diversity during the course of fermentation. As shown in Fig. 1A, the samples did not form very distinct clusters in the PCA coordinate space, indicating that the bacterial communities were in a state of continuous succession throughout the fermentation process. Moreover, the samples were more dispersed in the early stages of fermentation, reflecting a period of greater structural changes, while they tended to cluster together toward the middle and late stages. In comparison, the succession of fungal communities was more clustered than that of bacterial communities (Fig. 1B). Moreover, the fungal communities clustered together more among samples taken in the later stages of fermentation, indicating that the changes in fungal communities were minimal and tended to stabilize in the later stages of fermentation (Fig. 1A-B).

Fig. 1.

Changes and diversity of microorganisms during fermentation process. (A) Fungal PCA diagram. (B) Bacterial PCA diagram. (C) Diversity Index of Fungi. (D) Bacterial diversity index chart.

3.2.2. Changes in microbial community diversity during fermentation

Microbial community species diversity is typically represented by the alpha diversity indices, including the Shannon index, Chao index, and Simpson index. These metrics can provide a comprehensive evaluation of the species richness and evenness of a community. A higher Shannon index indicates higher species diversity, whereas a higher Chao index suggests higher species richness (Grice et al., 2009). When the coverage of all samples is greater than 90%, it is considered that the sequencing results can truly reflect the species abundance and diversity of the samples. As fermentation progressed from 0 to 14 days, fungal richness and diversity exhibited a progressive decline, whereas bacterial richness and diversity demonstrated an initial increase followed by a subsequent reduction. However, as the fermentation progressed, the diversity of the fungal community associated with alcohol showed a gradual downward trend (Fig. 1C-D). This phenomenon may be attributed to the coexistence of multiple microbial species in the fermentation environment. The proliferation of yeast species during fermentation generates ethanol through their metabolic activities (Gong et al., 2022). Consequently, the gradual increase in alcohol concentration resulting from ethanol accumulation creates a selective pressure that inhibits the growth of other microbial populations, thereby establishing yeast dominance during the fermentation process.

3.2.3. Dynamic changes in the microbial community during fermentation

The dynamic change trends of fungal and bacterial communities during the fermentation process were evaluated by tracking the main taxonomic groups (top 10) present in samples collected at different time points. The stacked column chart in Fig. 2 shows the relative abundances of identified bacterial and fungal taxa during the fermentation process. Fungal composition remained relatively homogeneous throughout fermentation. Three key fungal phyla were identified during the natural fermentation of pomelo wine. Ascomycota (74.74–95.63%) dominated as the predominant fungal phylum, maintaining absolute dominance throughout fermentation. Saccharomyces emerged as the primary fungal genus, accompanied by other genera such as Penicillium, Rhizopus, Komagataella, Aspergillus, and Wickerhamomyces (Fig. 2A-B). In the initial fermentation stage, Penicillium (45.17–0.42%) and Saccharomyces (24.23–81.52%) dominated, with yeast gradually becoming more dominant as fermentation progressed, with Saccharomyces replacing Penicillium as the dominant genus. This finding is in line with previous studies on fruit wine fermentation showing that Saccharomyces competes with other microorganisms during the fermentation process, ultimately establishing as the dominant bacterial genus (Wei et al., 2022; Perrone, Giacosa, Rolle, Cocolin, & Rantsiou, 2013).

Fig. 2.

The succession of microbial communities. (A) Succession of fungal colonies at the phylum level. (B)Succession of fungal colonies at the genus level. (C) Bacterial colony succession at the door level. (D) Bacterial colony succession at the genus level.

At the phylum level, the bacterial community was primarily dominated by Proteobacteria and Firmicutes. Proteobacteria exhibited a minimum relative abundance of 65.9% during fermentation, establishing it as the predominant phylum throughout the process (Fig. 2C). Previous studies suggested that Proteobacteria play a significant role in the early-stage microbial community of fermentation, where their metabolic activities may influence both the pH of the fermentation environment and the formation of flavor compounds (Barata, Malfeito-Ferreira, & Loureiro, 2012). At the genus level, the key bacterial taxa included Tatumella, Weissella, Bacillus, Gluconobacter, and Lactiplantibacillus, with Tatumella and Weissella identified as the dominant bacterial genera throughout fermentation. Notably, the relative abundance of these two genera exhibited a continuous change during the fermentation process, showing a trend of first increasing and then decreasing. At the end of fermentation, the relative abundance of Tatumella and Weissella reached 41.92% and 35.60%, respectively (Fig. 2D). Among the top five dominant genera, only Lactiplantibacillus was not detected at the beginning of fermentation (day 0), and then it showed an increase in relative abundance from day 1 to 7 of fermentation, ranging from 0.65% to 5.91% (Fig. 2D). Lactiplantibacillus can degrade acetaldehyde, and these species can also reduce the level of sulfur dioxide added in the wine-making process, even when it is combined with sulfur dioxide (Virdis, Sumby, Bartowsky, & Jiranek, 2021). Furthermore, Weissella and Lactiplantibacillus collaboratively participate in flavor metabolism (Yang et al., 2021). For example, Weissella can utilize carbohydrates to produce acetic acid and lactic acid, while Lactiplantibacillus plantarum primarily generates lactic acid via the glycolytic pathway. These organic acids provide the foundational sourness flavor of fermented products and further engage in esterification reactions to produce ester compounds, imparting fruity and sweet aromas (Xiong et al., 2024).

3.3. Hub taxa among the microbial community during pomelo wine fermentation

The network relationship diagrams of bacteria and fungi were constructed based on a threshold correlation of R > 0.8 and P < 0.05. According to the connectivity degree size ranking, the following taxa with ≥4 nodes were identified as hub microbial taxa during pomelo wine fermentation: Latilactobacillus, Oceanobacillus, Desemzia, Peniophora, Komagataeibacter, Kazachstania, Strelitziana, Kockovaella, Rubrobacter and Thermoactinomyces (Fig. 3). In the bacterial community, Latilactobacillus, Oceanobacillus, and Desemzia exhibited relatively good network connectivity (connectivity of 5), indicating that they may interact with multiple microorganisms during the fermentation process of pomelo wine. Latilactobacillus belongs to the family of lactic acid bacteria (LAB). Studies have found that Latilactobacillus reduces pH and inhibits the growth of harmful bacteria by producing lactic acid, thereby providing a safe fermentation environment for other microorganisms (Krieger-Weber, Heras, & Suarez, 2020) and promoting fermentation through mutual interaction. Oceanobacillus may possess some protease activity and exhibits cell wall-degrading enzyme activity, according to a whole-genome analysis of Oceanobacillus sp. SE10311 (Liu et al., 2025; Tan et al., 2024). Therefore, Oceanobacillus may promote fermentation in pomelo wine by breaking down certain proteins and other macromolecules to provide available substrates for other microorganisms, thereby forming interactions to facilitate fermentation. Desemzia is a relatively newly identified bacterial genus, and therefore understanding its role in fermented foods and beverages is currently very limited. Therefore, further studies are needed to clarify the existence, abundance, metabolic characteristics, and impact of Desemzia on the flavor, quality, and fermentation process of wine. Although its direct role is not yet clear, as a part of the microbial community, Desemzia may affect the fermentation environment and the flavor of the final product through interactions with other microorganisms. Overall, the dynamic succession of microbial communities and the ecological interactions between microorganisms jointly drive the orderly progress of the fermentation process.

Fig. 3.

Co-occurrence network of bacteria and fungi. Fungi (red nodes) and bacteria (green nodes) showed positive (red) and negative (blue) correlations. Node size indicates connectivity strength—larger nodes signify higher connectivity. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.4. Correlations of the microbial community and physicochemical indicators

Studies have shown that the metabolic activity, population dynamics, and succession process of the hub microbial community have a decisive influence on the changes of key parameters during wine fermentation (e.g., pH value, TS, TA, TSS, alcohol content). RDA was used to analyze the correlations between the hub microbial taxa and physicochemical indicators at different fermentation stages (Fig. 4). The first axis (RDA1) explained 66.99% of the variation, while the second axis (RDA2) explained only 8.17% of the variation.RDA1 is the dominant axis, reflecting physicochemical indicators (mainly pH and ethanol concentration) as the core driving factors shaping microbial community structure (Fig. 4).

Fig. 4.

Correlation between core microbiota and oenological parameters. RDA revealed the relationship between key microbial communities and their corresponding physicochemical characteristics. Red arrows denote the different oenological parameter variables, and blue arrows denote the hub taxa microbiota (Latilactobacillus, Oceanobacillus, Desemzia, Peniophora, Komagataeibacter, Kazachstania, Strelitziana, Kockovaella, Rubrobacter and Thermoactinomyces). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Specifically, the genera Strelitziana and Komagataeibacter were significantly positively correlated with the pH value and TSS during the fermentation process, whereas these taxa were only weakly correlated with TS, TA, and alcohol content. Desemzia, Latilactobacillus, Peniophora, Rubrobacter, and Oceanobacillus exhibited positive correlations with TS, whereas their correlations with pH, alcohol content, TA and TSS were relatively low. These results suggest that during brewing fermentation, these bacteria interact with the sugars present in the fermentation environment through different metabolic mechanisms.

The relative abundance of Thermoactinomyces was significantly correlated with the pH during the fermentation process. This suggests that Thermoactinomyces can affect the pH value of the fermentation broth through its metabolic activities. Studies have shown that Thermoactinomyces can produce organic acids (such as lactic acid and acetic acid) during the fermentation process. These acids can reduce the pH value of the fermentation broth, thereby creating a more acidic environment (Chen, Yang, Liu, Luo, & Zou, 2022). This acidic environment can then inhibit the growth of harmful microorganisms while promoting the metabolic activities of other beneficial microorganisms (such as yeast). Among them, Kazachstania and Kockovaella showed a significant positive correlation with TA and alcohol. Some studies have shown that certain species of Kazachstania exhibit significant ethanol production capacity during fermentation, producing relatively high concentrations of higher alcohols (Jood, Hoff, & Setati, 2017). Therefore, the dynamic changes of hub taxa microorganisms during the fermentation process enhance their adaptability to exert changes to the fermentation environment. In this process, these microorganisms also produce specific flavor compounds, which affect the flavor and alcoholic sensation of the final pomelo wine product (Peng et al., 2021).

3.5. Dynamic changes in volatile compounds during pomelo wine fermentation

A total of 2016 volatile compounds were identified during the fermentation process of pomelo wine, including esters, terpenes, ketones, heterocyclic compounds, alcohols, acids, and aldehyde as the main categories (Table S1, Fig. 5A). PCA of all metabolites revealed a clear gradient distribution among samples collected at different time points of fermentation along the PC1 axis. The close distribution of three sample points in 0d along PC1 and PC2 indicates high consistency in the metabolite composition during the initial stage of fermentation. From GZ_1d to GZ_14d, although there was slight dispersion among the three replicate samples at each time point, they still showed a clear time gradient along the PC1 axis, indicating good intra-group repeatability and reliable results (Fig. 5B). In addition, PCA of the metabolites identified between GZ_0d and GZ_14d showed that the sample point of GZ_0d clustered in the positive direction of PC1 (to the right of the horizontal axis), while the sample point of GZ_14d clustered in the negative direction of PC1 (to the left of the horizontal axis), and the two groups showed significant separation on the PC1 axis. This indicates that after 14 days of fermentation, the metabolite composition of pomelo wine underwent significant dynamic changes (Fig. 5C). The relative content of alcohols increased at the fastest rate, with a 7.128-fold increase at 14 days (18.96%) compared to the content measured at 0 days (2.66%) (Fig. 5D). Alcohols are precursors of aged esters, and the key enzyme involved in alcohol production is alcohol dehydrogenase (ADH), which catalyzes the conversion of sugars into alcohol (Qian et al., 2019). Yeasts (such as Saccharomyces cerevisiae) are the key microorganisms playing a role in alcoholic fermentation, which synthesize higher alcohols through the Ehrlich pathway and amino acid metabolic pathways. The key enzymes in these metabolic pathways (such as 2-keto acid decarboxylase and alcohol dehydrogenase) continuously catalyze reactions during the fermentation process, leading to the overall accumulation of alcohol substances.

Fig. 5.

Dynamic changes in volatile compounds during Citrus wine fermentation. (A)Volatile compound category pie chart. (B) PCA score scatter plot (all). (C)PCA score plot comparing fermentation stages (GR-0d vs. GR-14d).  (D) Changes in volatile compound content from 0d to 14d during fermentation process.(E) Contribution changes of 15 metabolites during fermentation from 0d to 14d (F) Classification and aroma contributions of 15 metabolites.

Alcohols play an important role in the flavor of fruit wine: they can react with acids to produce aromatic substances, reduce the pungency of fruit wine, and enhance its harmony (Yang, Luo, & Wang, 2014). The relative content of ester metabolites increased from 10.49% on day 0 of fermentation to 20.18% on day 14 (Fig. 5D). This increase was mainly related to the activity of the acyltransferase enzyme (Chen, Steinhauer, Hammerlindl, Keller, & Zou, 2007).

After screening the metabolites detected in the sample according to a variance importance in projection (VIP) value >1, fold change >2 or < 0.5, and P < 0.05, 315 metabolites were identified (Table S2). As the fermentation progressed, pomelo wine exhibited an increase in esters (fold change >200, VIP > 1) such as 3-hexenoic acid, methyl ester, (Z)-; p-menth-8-en-3-ol, acetate; pentanoic acid, phenylmethyl ester; cyclohexanol, 1-methyl-4-(1-methylethylidene)-, acetate; butanoic acid, 2-methyl-, phenylmethyl ester; 2,6-octadienoic acid, 3,7-dimethyl-, methyl ester; 2-butenoic acid, 2-methyl-, ethyl ester; 5-tetradecen-1-ol, acetate, (Z)-; and 11-dodecen-1-yl acetate, with 3-hexenoic acid, methyl ester, (Z)- being the most abundant volatile compound in the category (Fig. 5C-D; Table S2). At day 0 of fermentation, the content of terpenoids was the highest, reaching 61.52%. As the fermentation process progressed, the relative content of terpenoids gradually decreased, becoming 22.2% on the 14th day of fermentation. This substantial decrease reflects the activities of Saccharomyces and other microorganisms such as Bacillus that utilize terpenes. Saccharomyces uses terpenes as an auxiliary carbon source to degrade monoterpenes through the β-oxidation pathway. Some strains of Bacillus secrete terpenoid-degrading enzymes (Xie et al., 2024). This biotransformation changes the composition and relative content of terpenoids, which in turn affects the aroma of the fruit wine (Roberts et al., 2024). The relative contents of ketone metabolites and heterocyclic compounds increased by 0.28% and 3.88%, respectively, during fermentation, but their overall percentage changes were not significant (Fig. 5C-D). It is possible that the main fermenting microorganisms in the brewing process are yeasts, and their metabolic activities are mainly concentrated on the decomposition of sugars and the production of alcohol. Yeasts tend to produce esters and alcohols during fermentation. The production of heterocyclic compounds usually requires specific precursor substances and metabolic pathways, which do not play a dominant role in the metabolic process of yeasts (Deng et al., 2025).

3.6. rOAV analysis of volatile compounds during the fermentation process of pomelo wine

rOAV analysis is a commonly used method to evaluate the contribution of different compounds to overall aroma composition. This value can help to determine the changes and relative importance of volatile compounds during the fermentation process. In general, a larger rOAV indicates a greater contribution of a given compound to the overall flavor (Zhang et al., 2019). In this study, aroma compounds were selected based on their odor characteristics and corresponding 315 rOAVs (Table S2). Among them, 15 compounds were considered as key differential aroma compounds, characterized by an rOAV exceeding 1000, VIP > 1, fold change >2 or < 0.5, and P < 0.05(Fig. 5E-D; Table S2). Among them, the following were the main contributing compounds to highlight the aroma of pomelo wine in terms of “grapefruit,” “sweet,” and “fruity” aromas (0 d): 3-cyclohexene-1-methanethiol, alpha,alpha,4-trimethyl- (rOAV = 4895.01); ethanone, 1-(2-aminophenyl)- (rOAV = 1418.57); decanoic acid, ethyl ester (rOAV = 6.63); phenylethyl alcohol (rOAV = 5.26); octanoic acid, ethyl ester (rOAV = 0.5); oxacycloheptadecan-2-one (rOAV = 5); 4-phenyl-2-butanol (rOAV = 1.22); and 3-mercaptohexyl acetate (rOAV = 28.6) (Fig. 5E-D; Table S2). During the fermentation process, environmental and microbial factors act together to cause changes in the aroma characteristics of pomelo wine. For example, the aromas of “grapefruit,” “sweet,” “grape,” and “fruity” become increasingly intense and blend with some fresh scents. The contributions of the aromatic characteristics “woody,” “roasted,” “mushroom,” and “oily” gradually decreased during the fermentation process. Additionally, the ethyl ester decanoic acid contributes not only to aromas such as “sweet” and “fruity” but is also associated with the “brandy” aroma. Notably, 1-octen-3-one mainly contributes to the “mushroom” flavor, and its rOAV was 788.21 on the 14th day of fermentation. The various aromas contributed by these volatile compounds enrich and enhance the sensory attributes of pomelo wine.

3.7. Correlations between the hub microbial taxa and key differential aromas

To determine the correlations between the hub taxa microbiota and key differential aromas, we constructed a connection network that met the conditions of Pearson correlation coefficient |R| ≥ 0.6 and P < 0.05 (Fig. 6). This network facilitates predicting the role of the microbiota in the formation of differential aromas during pomelo wine fermentation. Bacillus, Pichia, Komagataeibacter, Pediococcus, Zasmidium, Penicillium, Saccharomyces, Weissella, and Gluconobacter demonstrated robust correlations with the key differential aromas. Among them, Komagataeibacter, Bacillus, Penicillium, Zasmidium, and Gluconobacter were all negatively correlated with various volatile aroma metabolites, and their relative abundances decreased with increasing fermentation time. Specifically, the relative abundance of Komagataeibacter changed from 0.084% to 0, Bacillus changed from 11.36% to 1.36%, Penicillium changed from 45.17% to 0.42%, Zasmidium changed from 0.82% to 0, and Gluconobacter changed from 1.98% to 0. All five of these microorganisms were associated with oxacycloheptadecan-2-one (“sweet” and “musky”)，3-cyclohexene-1-methanethiol,alpha,alpha,4-trimethyl- (“sulfury,” “aromatic,” and “grapefruit”), ethanone,1-(2-aminophenyl)- (“grape” and “sweet”), and trans,cis-2,6-nonadien-1-ol (“green” and “cucumber”). The volatile aroma metabolites exhibited negative correlations with the microorganisms. Among them, Penicillium, as a fungus, may compete for carbohydrate substrates, which may affect yeast ethanol fermentation and alter metabolite composition, potentially affecting ester synthesis. At the same time, the relative abundance Penicillium decreased due to competition failure (from 45.17% to 0.42%), and this change was negatively correlated with the change in metabolites (Dartora et al., 2023). Komagataeibacter may increase the oxidation-reduction potential by oxidizing ethanol to acetic acid, leading to the decomposition of certain easily oxidizable metabolites such as trans,cis-2,6-nonadien-1-ol, which disappeared by the end of fermentation (from 0.084% to 0), likely due to high oxidative environment inhibition (Ohwofasa, Dhami, Winefield, & On, 2024). Zasmidium is a fungal genus that may compete with yeast for nitrogen sources. When the abundance of Zasmidium decreases, yeast has access to more nitrogen sources, promoting ester synthesis (such as derivatives of ethanone,1-(2-aminophenyl)-), which results in a further decrease of Zasmidium due to nitrogen source limitation and is negatively correlated with key aroma metabolites.

Fig. 6.

The relationship network between hub taxa microbiota and key differential aroma compounds (Pearson; R > 0.6, P < 0.05). The red nodes display key differential aroma compounds, while the blue nodes display the core microbial community. The pink and green lines indicate correlation, with green indicating negative correlation and pink indicating positive correlation. Node size reflects connectivity. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Weissella, Saccharomyces, Pichia, and Pediococcus were positively correlated with a variety of volatile aroma metabolites. The relative abundances of these four microorganisms showed an increasing trend during the fermentation process, with that of Weissella changing from 8.20% to 16.88%, Saccharomyces changing from 24.22% to 81.52%, Pichia changing from 0.02% to 0.63%, and Pediococcus changing from 0.12% to 1.35%. In addition, these four microorganisms were all positively correlated with phenylethyl alcohol (“fruity” and “sweet”), ethanone,1-(2-aminophenyl)- (“grape” and “sweet”), decanoic acid, ethyl ester (“sweet” and “fruity”), oxacycloheptadecan-2-one (“sweet” and “musky”), and pyrazine,2,3-diethyl-5-methyl- (“nut skin” and “sweet”). Saccharomyces produces a potential precursor for ethanol, ethanone,1- (2-aminophenyl)-, through ethanol fermentation, which participates in the reaction. At the same time, its esterase activity may catalyze the synthesis of decanoic acid and ethyl ester. When the abundance of Saccharomyces increases, the contents of ethanol and ester precursors increase in turn, indirectly promoting metabolite accumulation. A previous study found that after the attenuation of Zasmidium, the content of ethanone, 1-(2-aminophenyl)- increases (Jansen et al., 2024). Weissella can lower the fermentation pH and consequently reduce the relative abundances of acid-resistant microorganisms such as Penicillium and Zasmidium. However, the acidic environment makes it easier for alcohols to diffuse from the cell membrane, which may promote the release of phenylethyl alcohol. Meanwhile, the Weissella abundance increased during fermentation due to its ability to adapt to low pH environments, forming an indirect association with metabolites (Ohwofasa et al., 2024).

The relative content of Saccharomyces increased from 24.22% on day 0 of fermentation to 81.52% on the 14th day of fermentation. At the end of fermentation, its relative content accounted for four-fifths of the total microorganisms, showing the greatest increase. Saccharomyces is the most important yeast genus in fruit wine fermentation, which usually dominates the fermentation process. This yeast genus can efficiently convert sugar into ethanol and carbon dioxide while producing a large number of volatile aroma compounds. Studies have shown that Saccharomyces can produce a variety of alcohols and esters, which are the main components of fruit wine aroma. Pichia can produce some special aroma compounds such as benzaldehyde, which has the aroma of bitter almonds and nuts. When co-cultured with Saccharomyces, Pichia can also inhibit the production of certain undesirable aroma compounds such as 4-methylphenol, thereby improving the overall aroma quality of wine (Ding, Zhao, Zhang, Lin, & Xiong, 2024). Therefore, the succession and joint action of these microorganisms, as well as changes in the fermentation environment, directly or indirectly affect the production of key volatile aroma compounds during the fermentation process, thereby enriching the flavor and quality of pomelo wine.

4. Conclusions

This study investigated the relationships among the dynamic changes in physicochemical parameters, volatile metabolites, and microbial community succession during pomelo wine fermentation through the analysis of microbial community dynamics and volatile metabolites. In the fungal kingdom, the yeast Saccharomyces is crucial in shaping the course of fermentation. Additionally, Latilactobacillus, Oceanobacillus, Desemzia, Peniophora, Komagataeibacter, Kazachstania, Strelitziana, Kockovaella, Rubrobacter, and Thermoactinomyces were identified as the hub genera in the bacterial fermentation community, which had significant correlations with changes in physicochemical parameters during pomelo wine fermentation. The dynamic changes in microorganisms directly affected the theoretical parameters during pomelo wine fermentation. Moreover, Saccharomyces and Weissella had positive correlations with a variety of key volatile aroma compounds, indicating that these genera play a key role in the formation of characteristic flavor compounds of pomelo wine. The correlation analysis of key discriminant aromas and microorganisms confirmed that microorganisms play an important role in the aroma formation during the fermentation of pomelo wine. These findings provide a deeper understanding of the production mechanism of flavored beverages, especially fruit wines, which can serve as a guide to select excellent strains with flavor-modifying effects, thereby achieving effective control of flavor quality in the industrial production of pomelo wine.

Glossary

ASV	Amplicon sequence variants
LAB	Lactic acid bacteria
PCoA	Principal coordinate analysis
PCR	Polymerase chain reaction
TA	Total acidity
TS	Total sugar
TSS	Total soluble solids
VIP	Variance importance in projection

CRediT authorship contribution statement

Yang Wu: Writing – review & editing, Software, Resources, Project administration. Zexia Li: Writing – review & editing, Writing – original draft, Methodology, Investigation, Formal analysis, Data curation. Xiujuan Duan: Investigation, Data curation. Liting Dong: Investigation, Data curation. Mingxue Chang: Investigation, Data curation. Wei Zheng: Methodology, Investigation. Yongfa Guo: Methodology, Investigation, Data curation. Weiwei Li: Methodology, Data curation. Dingkun Liu: Writing – review & editing. Hai Yu: Methodology, Investigation. Huimin Sun: Writing – review & editing, Supervision, Resources, Project administration, Funding acquisition.

Funding

This work was supported by the National Natural Science Foundation of China (31860366), Jiangxi Provincial Natural Science Foundation (20192BAB204013), Ji’an City Science and Technology Plan Project (20244-018566 and 20244-018557), Key Laboratory of Jiangxi Province for Biological Invasion and Biosecurity (2023SSY02111), Key Laboratory of Jiangxi Province for Functional Biology and Pollution Control in Red Soil Regions (2023SSY02051) and National and Provincial College Students' Innovation and Entrepreneurship Training Program (202510419020).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material 1

Supplementary material 2

Data availability

Data will be made available on request.