Multi-omics study on traditional Xiaoqu of Fangxian Huangjiu: Unveiling interactions among physicochemical factors, microbiome and volatile flavor compounds

Abstract

Xiaoqu is the key fermentation starter of Huangjiu, with its microbial profile closely linking to the flavor quality. This study investigated Fangxian Xiaoqu (FXXQ) and corresponding Fangxian Huangjiu (FXHJ) by physicochemical analysis, microbiomics, and flavoromics. Results showed significant variations in physicochemical properties, with Weissella and Saccharomycopsis identified as dominant bacteria and fungi, respectively. A total of 255 bacterial and 151 fungal differential taxa in FXXQ were identified, as well as 76 volatile flavor compounds in FXHJ, including ten newly reported for FXHJ. Correlation analysis revealed acid protease activity as the most critical physicochemical factor shaping bacterial communities, while alcohol-producing capacity dominated fungal communities. Esters content was positively correlated with the abundances of Leuconostoc and Staphylococcus, and alcohols content with that of Saccharomycopsis. This study provides theoretical support for standardizing the preparation of FXXQ and regulating the flavor of the final FXHJ product.

Highlights

• •A multi-omics approach was first applied to probe typical Fangxian Xiaoqu (FXXQ).
• •Key physicochemical indicators of fermentation were systematically quantified.
• •255 bacterial and 151 fungal differential taxa served as potential biomarkers.
• •Ten compounds were newly reported in Fangxian Huangjiu.
• •Core microbes showed key correlations with physicochemical and flavor profiles.

1. Introduction

Huangjiu, one of the three ancient wines in the world, is brewed from grains such as rice and millet through simultaneous saccharification and fermentation. This unique brewing process is dependent on Qu, a complex starter culture colonized by amylolytic fungi, fermentative yeasts, and lactic acid bacteria (Peng, Zheng, Meng, Yu, et al., 2022; Peng et al., 2023; Sun, 2019). This microbial consortium collectively governs starch hydrolysis, alcohol production, and flavor metabolism (Peng et al., 2023; Peng, Zheng, Meng, Zhu, et al., 2022; Xie et al., 2021), forming a diversified regional system based on raw material and starter differences (Lv et al., 2013). Fangxian Huangjiu (FXHJ) holds national geographical indication protection and provincial intangible cultural heritage status (Chen, 2019). It is rich in amino acids, peptides and polyphenols, and characterized by high β-phenylethanol and low isoamyl alcohol contents. Its brewing features simultaneous saccharification and fermentation driven by complex microbial system of local Xiaoqu, yielding a creamy white to pale yellow product with prominent rice fragrance and unique flavor (Zhu et al., 2018).

As a saccharification and fermentation starter, Qu is an important carrier of microbial resources in the brewing process (Wang et al., 2021). In Huangjiu production, the commonly used Qu mainly includes three types: wheat Qu, Hongqu and Xiaoqu (Xia et al., 2024). Wheat Qu serves as the core saccharification and fermentation agent in Huangjiu production, with advantages of short production cycles, suitability for mechanized manufacturing, and high product quality stability, though it has relatively low microbial diversity (Ma et al., 2024). Hongqu endows fermented products with a distinctive bright-red color, unique taste, and special flavor, along with relatively stable microbial diversity. However, its production involves multiple complex steps including rice soaking, steaming, inoculation, sprouting, floor cultivation, propagation, and sun-drying (Xia et al., 2024). Located in northwestern Hubei Province, Fangxian specializes in lump or bolus Xiaoqu, formulated from japonica rice, Polygonum hydropiper L., Chinese medicinal herbs, and the seed Qu produced in the previous years. These ingredients are combined through scientific proportioning and appropriate water addition by traditional craftsmanship. Fangxian Xiaoqu (FXXQ) has advantages of simple production, low usage, and easy storage (Zhang et al., 2022; Zheng et al., 2023b), but its traditional and experiential preparation results in uncontrollable fermentation and inconsistent quality. While studies have focused on physicochemical parameters and enzymatic activities for process control (Chen et al., 2020; Zhang, Jia, et al., 2025; Zhao et al., 2023; Zhu et al., 2023), research related to FXXQ remains preliminary and lacks a comprehensive analytical framework (Zhang, Lin, et al., 2025), hindering standardized production and quality control.

The microbial community of Xiaoqu mainly comprises molds, yeasts and bacteria (Kang et al., 2022). Advances in high-throughput sequencing (HTS) technology have facilitated in-depth investigation of microbial diversity. Zhang et al. (2022) characterized the microbial community of 6 types of FXXQ, revealing 15 dominant bacterial genera (e.g. Weissella and Pediococcus), and 13 dominant fungal genera (e.g. Wickerhamomyces and Rhizopus). These microorganisms produce diverse enzymes facilitating saccharification, alcoholic fermentation, and flavor biosynthesis (Yang et al., 2017). Molds serve as primary saccharification agents, producing enzymes such as glucoamylase, α-amylase and acid protease, which hydrolyze starch and protein substrates in raw materials while generating flavor compounds (Wang, Chen, et al., 2024). Yeasts dominate ethanol biosynthesis, whereas bacteria exhibit superior capabilities in aroma compounds production. These metabolic functions arise from enzymes synthesized by microbial enzymes (Yang et al., 2020). Gao et al. (2024) identified 18 starch-metabolizing enzymes and 71 flavor-associated enzymes in the FXHJ fermentation system. Specifically, α-amylase derives from the synergistic expression of Rhizopus, Saccharomycopsis, and Lactococcus, while the flavor network features n-propanol, 2-methyl-1-butanol and isobutanol. This multimicrobial enzyme system constitutes the biochemical basis for the formation of the unique flavor of FXHJ.

Flavor is one of the key factors determining the quality and style of Huangjiu. Headspace solid-phase microextraction (HS-SPME) combined with gas chromatography–mass spectrometry (GC–MS) has been widely applied for its high efficiency and sensitivity (Peng, Zheng, Meng, Yu, et al., 2022; Wang, Chen, et al., 2024). Specifically for FXHJ, 28 substances have been identified as characteristic flavor compounds, including 9 esters, 2 alcohols and 5 acids and other key components (Gao, Peng, et al., 2024). However, current research on FXHJ and FXXQ lacks integrated multi-omics analyses, leaving the microbial community structure and function remaining poorly characterized. Previous studies focused on isolated aspects without systematic integration. This limitation hinders process optimization and quality control in actual Huangjiu production.

Thus, this study aimed to systematically characterize FXXQ and its role in FXHJ quality formation. Specific objectives included: (1) quantify the key physicochemical indices of 12 representative varieties and their resultant FXHJ to clarify the foundational quality traits of FXXQ and its brewing potential; (2) analyze the microbial community structure of FXXQ via HTS and identify biomarker taxa using Linear discriminant analysis Effect Size (LEfSe); (3) predict and annotate microbial functions based on Non-supervised Orthologous Groups (EggNOG) and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases; (4) quantify the VFCs of FXHJ through HS-SPME and GC–MS to define its characteristic flavor profile; and (5) establish multivariate correlations between dominant microbes, physicochemical indexes and VFC to reveal the functional mechanisms of FXXQ in flavor formations. The findings of this research will advance traditional brewing modernization and provide a scientific basis for preserving regional fermentation heritage.

2. Materials and methods

2.1. Materials and chemical reagents

Glutinous rice was purchased from Shenxin Rice Co., Ltd. (Changzhou, China). Twelve representative Xiaoqu samples were produced in July 2024 and collected from Fangxian in Shiyan City, Hubei Province, China (Fig. S1). CS (Caoshi) was supplied by Cao Family Jiuqu Factory; DS-DF (Dengshi-dafang), DS-XF (Dengshi-xiaofang) and DS-Y (Dengshi-yuan) originated from Deng Family Qu Industry; HQ-BT (Huiquan-bantian) and HQ-T (Huiquan-tian) were collected from Huiquan Huangjiu Factory; JQ (Jianqing) originated from Fanglu Shenniang Brewery; LFQ (Lefoquan) was supplied by Lefoquan Qu Factory; SJ-T (Sunjun-tian) and SJ-YJ (Sunjun-youjin) came from Sunjun Premium Qu Factory; YN-K (Yunneng-ku) and YN-T (Yunneng-tian) were provided by Yunneng Qu Factory. The geographical locations of these factories are listed in Table S1, and the geographic distances between factories are shown in Table S2. These samples encompassed the principal producers within Fangxian and are highly representative. All the samples were preserved at 4 °C for further analysis.

Absolute ethanol (99.9%), sodium chloride (99.8%), hydrochloric acid (36.0–38.0%), copper sulfate pentahydrate (99.0%), potassium sodium tartrate (99.0%), sodium hydroxide (98.0%), glucose (99.0%), methylene blue (96.0%), methyl red (95.0%), sodium acetate anhydrous (99.0%), sodium carbonate anhydrous (99.8%), trichloroacetic acid (99.0%) and disodium hydrogen phosphate (99.0%) were purchased from Sinopharm Chemical Reagent Co., Ltd. (Beijing, China). Acetic acid (99.7%), L-tyrosine (99.0%), citric acid (98.0%) and 4-octanol (98.0%) were purchased from J&K Scientific Co., Ltd. (Beijing, China). Dilute iodine solution and soluble starch were purchased from Biorigin Biotechnology Co., Ltd. (Beijing, China). OED60K type fermentation industry universal foaming agent was purchased from Jinlis Technology Co., Ltd. (Beijing, China). Folin-Ciocalteu reagent was acquired from Solarbio Technology Co., Ltd. (Beijing, China). The lipase activity kit was purchased from Geruisi Biotechnology Co., Ltd. (Suzhou, China). DL-Lactic acid (80.0–85.0%) was purchased from Innochem Technology Co., Ltd. (Beijing, China). Sodium DL-lactate solution (73.0%) was purchased from Aladdin Biochemical Technology Co., Ltd. (Shanghai, China). Casein was obtained from National Institutes for Food and Drug Control (Beijing, China).

2.2. Fermentation experiments

The steps were as follows: First, 12 portions of glutinous rice (500 g each) were washed three times, soaked in water for 15 h, and steamed for 15 min. Then each portion was transferred to an individual sterile 2.5 L earthenware jar and cooled to 30 °C with the addition of 300 mL of sterile water. Subsequently, 3.00 g of the corresponding FXXQ, pre-crushed by a pulverizer, was added to each jar, and the mixture was stirred thoroughly. The jars were covered with gauze, labeled, and placed in a constant temperature incubator at 28 °C for a 7-day pre-fermentation. Post-fermentation was then carried out at 17 °C, with a total fermentation period of 30 days (Yang et al., 2020; Zheng et al., 2025). Finally, 100 mL of each Huangjiu sample was collected in a sterile centrifuge tube and stored at −80 °C for subsequent analysis.

2.3. Determination of physicochemical indexes

The determination of moisture content, acidity, saccharification power and alcohol-producing capacity of FXXQ was referred to QB/T 4257–2011 General methods of analysis for Daqu. Acid protease activity was determined by Folin method following GB/T 23527.1–2023 Quality requirements for enzyme preparations-Part 1: Protease preparations, while α-Amylase activity was determined according to GB/T 24401–2009 Alpha-amylase preparation. Lipase activity was determined following the instructions provided by Geruisi Biotechnology Co., Ltd. Briefly, 0.1 g of sample was homogenized with 1 mL extraction solution on ice. The homogenate was centrifuged at 12,000 rpm and 4 °C for 10 min. Subsequently, 10 μL of supernatant was added to a 96-well plate, followed by 40 μL of Reagent I and 150 μL of Reagent II. After mixing, the initial absorbance was measured immediately at 405 nm and 30 °C. After a 10-min incubation under the same conditions, the secondary absorbance was measured. Lipase activity was calculated using formula 1. The physicochemical indexes for each sample were determined in triplicate.

$$\mathit{LPS}=\frac{307.7\times \left(A_2-A_1+0.0003\right)}{W}$$ (1)

LPS: Lipase activity. The unit is nmol/min/g fresh weight;

A₂: The secondary absorbance reading;

A₁: The initial absorbance reading;

W: The exact weight of the sample. The unit is g.

The pH, total acid content, and total sugar content of FXHJ were determined according to GB/T 13662–2018 Huangjiu, with pH measured using a calibrated pH meter (PH-100, Shanghai Lichen Instrument Technology Co., China). Alcohol content was determined by 1260 high-performance liquid chromatography (HPLC, Agilent Technologies, USA). Huangjiu samples were filtered through 0.45 μm polyethersulfone membranes (Baoruyi Biotechnology Co., Ltd., China) before injection. The HPLC conditions were as follows: the mobile phase was ultrapure water at a flow rate of 0.4 mL/min, with RID detection at 55 °C. The chromatographic column was Hi Plex H (300 × 7.7 mm, 8 μm; Agilent Technologies, USA), and the column oven temperature was 55 °C. Quantification was based on an external calibration curve prepared with ethanol standard solutions at concentrations of 1%, 2%, 4%, 8%, 12%, 16% and 32%. The physicochemical indexes were analyzed in triplicate.

2.4. DNA extraction, PCR amplification and sequencing

DNA extraction and PCR amplification were performed as described in previous literature (Ren et al., 2019). For bacterial 16S rRNA gene analysis, the V3-V4 region was amplified using primers 338F (5′-ACTCCTACGGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) under the following conditions: initial denaturation at 95 °C for 3 min; 27 cycles of denaturation (95 °C, 30 s), annealing (55 °C, 30 s), and extension (72 °C, 45 s); final extension at 72 °C for 10 min. Fungal ITS regions were targeted with primers ITS1F (5′-ACTTGGTCATTTAGAGGAAGTAA-3′) and ITS2R (5′-GCTGCGTTCTTCATCGATGC-3′) using 35 cycles of the same thermal parameters. The PCR products were sequenced by Majorbio Co., Ltd. (Shanghai, China) at the MiSeq platform of an Illumina high-throughput sequencer (Illumina, USA). All analyses were repeated three times.

2.5. Characterization of VFCs of FXHJ

The VFCs of FXHJ were extracted by headspace solid-phase microextraction (HS-SPME) and analyzed using a TRACE1610 gas chromatography (Thermo Scientific, USA) equipped with an Orbitrap Exploris GC mass spectrometer (Thermo Scientific, USA). For extraction, 2 g of NaCl, 5 mL of sample, and 10 μL of the internal standard mixture (4-octanol, 500 mg/L) were added to a 20 mL headspace vial. Volatile compounds were adsorbed using a 50/30 μm DVB/C-WR/PDMS SPME fiber (Thermo Scientific, USA). The separation was separated on a DB-FFAP column (30 m × 0.25 mm × 0.25 μm, Agilent, USA). The injection and GC–MS conditions were described in a previous report (Zheng et al., 2023b). Qualitative analyses were achieved by the NIST17 (National Institute of Standards and Technology) spectral library (match >700) and retention index (RI) (Ye et al., 2022). Under the same chromatographic conditions, the RI of VFCs calculated using C7-C27 n-alkanes (Supelco, USA) were compared with those previously reported, where the absolute values differing by 50 or less were determined to be the same compound (Wang et al., 2022). An internal standard method was used for the quantification of the VFCs (Yang et al., 2022). All experiments were repeated in triplicate.

2.6. Bioinformatics and statistical analysis

Raw sequencing reads were processed and clustered into operational taxonomic units (OTUs) with Usearch (version 11). Microbial α-diversity was analyzed via mothur software (version 1.30.2). Venn diagrams were generated using the VennDiagram package in R Studio (version 4.5.0). β-diversity was quantified through Bray-Curtis distance matrices in QIIME (version 1.9.1). Nonmetric multidimensional scaling (NMDS) was conducted and visualized with the vegan package in R. LEfSe was employed to identify significantly different species among groups (LDA > 3.5) via the microeco package in R. Functional predictions of the microbial communities were performed using PICRUSt (version 1.1.0) within the R environment. Canonical correspondence analysis (CCA) and redundancy Analysis (RDA) were performed to explore the relationships between physicochemical properties and microorganisms, with results visualized by the vegan package in R. Spearman correlation analysis was performed between microbial genera (relative abundance >1%) and VFCs (content >10 μg/L), and the results were visualized as a heatmap in R Studio.

Data processing and analysis were performed using Microsoft Office Excel 2016 and SPSS Statistics 24. Differences were considered statistically significant at p < 0.05. Principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were conducted using Simca (version 14.1). Visualization of physicochemical indicators and VFCs contents was performed using GraphPad Prism (version 10.1.2) and Origin 2022, respectively.

3. Results and discussion

3.1. Physicochemical indexes of FXXQ and FXHJ

The physicochemical properties of FXXQ are shown in Fig. 1 and Table S3. Moisture content during Qu fermentation critically influences microbial growth and metabolism, thereby indirectly impacting flavor formation (Sun et al., 2023). Significant variations in moisture content were observed among FXXQ from different manufacturers, ranging from 8.99% to 17.07%. Notably, substantial differences also existed within the same manufacturer, as exemplified by HQ-BT and HQ-T, indicating distinctiveness of their production processes. This variation may directly affect the activity of microorganisms like lactic acid bacteria (LAB) within the koji, potentially influencing subsequent fermentation (Tian et al., 2025). As a composite microbial agent integrating saccharification, fermentation and aroma production, Qu generates various organic acids through metabolic processes (Zhang & Chang, 2009). The acidity of samples ranged from 0.09 to 0.37 mmol/10 g. Saccharification, the enzymatic hydrolysis of complex carbohydrates into monosaccharides, contributes to flavor development during fermentation (Huang et al., 2024). As a key quality indicator for Qu, saccharification power varied from 98.60 to 317.00 U among the samples, with DS-Y and DS-DF exhibiting significantly higher saccharification power (above 300 U) than other samples.

Fig. 1.

Analysis of physicochemical indexes and enzyme activity of Fangxian Xiaoqu (FXXQ). (a) moisture content; (b) acidity; (c) saccharification power; (d) acid protease activity; (e) α-amylase activity; (f) lipase activity; (g) alcohol-producing capacity. CS, DS-DF, DS-XF, DS-Y, HQ-BT, HQ-T, JQ, LFQ, SJ-T, SJ-YJ, YN-K and YN-T were represented Caoshi, Dengshi-dafang, Dengshi-xiaofang, Dengshi-yuan, Huiquan-bantian, Huiquan-tian, Jianqing, Lefoquan, Sunjun-tian, Sunjun-youjin, Yunneng-ku and Yunneng-tian Xiaoqu, respectively. Significant differences are indicated by different letters (p < 0.05).

Acid proteases play a significant role in the formation of flavor precursors in Huangjiu (Yu et al., 2015). The acid protease activity of FXXQ samples ranged from 9.64 to 330.97 U/g. CS, DS-DF, DS-XF and JQ exceeded 100 U/g, significantly higher than the others. This high activity may be related to the high abundance of LAB such as Pediococcus (Jin et al., 2017). Amylase hydrolyses starch in grains to provide nutrients for microbial growth, and its activity directly influences the fermentation rate and ethanol yield, making it a key factor in determining Huangjiu quality (Liu et al., 2020). The α-amylase activity was 4.55–4.75 U/g. Lipase activity varied considerably among samples, ranging from 51.48 to 224.52 nmol/min/g fresh weight. YN-K exhibited the highest lipase activity, which may be linked to its notably high abundance of Wickerhamomyces (Shimizu et al., 2020). Alcohol-producing capacity, which integrates liquefaction power and saccharification fermentation power, directly reflects the wine-production capacity of Qu (Yuan et al., 2017). This index varied among samples, with values ranging from 7.80 to 13.74%vol.

Key physicochemical parameters of FXHJ were determined (Table S4). The pH ranged from 3.62 to 4.12, and total acid content varied between 9.26 and 14.58 g/L, with DS-DF, DS-XF, DS-Y and YN-K showing significantly lower total acid levels than other samples. The total sugar content ranged from 57.63 to 256.73 g/L. With total sugar contents ranging from 40.1 to 100.0 g/L, HQ-BT, JQ and LFQ were semi-sweet Huangjiu as defined by GB/T 13662–2018. Other samples exhibiting total sugar contents above 100.0 g/L were sweet Huangjiu. DS-DF and DS-XF showed significantly higher sugar content than others. The alcohol content of FXHJ varied widely (5.46–12.26%vol), indicating considerable variation in fermentation efficiency and alcohol yield across samples (Zhu et al., 2023). Specifically, HQ-T, JQ, LFQ, SJ-T, SJ-YJ, YN-K, and HQ-BT had the highest alcohol content with no significant differences among them, whereas the lowest levels were found in DS-DF, DS-XF, and YN-T. Compared with Shanxi broomcorn millet Huangjiu (SXHJ) (Li, Guo, et al., 2025) and GY from Shaoxing (Luo et al., 2024), FXHJ exhibited significantly higher total acidity and total sugar content, with an alcohol content lower than that of SXHJ but higher than that of GY.

3.2. Microbial community analysis of FXXQ

3.2.1. Sequence analysis of FXXQ samples

The bacterial and fungal community diversity of FXXQ was analyzed. For bacterial communities, 1596096 valid sequences were obtained with an average sequence length of 425 bp. A total of 377 OTUs were clustered at 97% similarity and labeled using the Silva database. For fungal communities, 1444284 valid sequences were obtained with an average sequence length of 263 bp, and 250 OTUs were clustered. The number of sequences and OTUs were higher for bacteria than for fungi.

3.2.2. Α-Diversity analysis of FXXQ microbial communities

Plateaued dilution curves (Fig. S2) of α-diversity confirmed adequate sequencing depth and near-complete microbial diversity, indicating the data reliably represents the sample microbial diversity.

The α-diversity of the microbial community was analyzed, including the coverage, richness and diversity of bacteria and fungi (Fig. 2 and Table S5–6). The coverage indices of bacteria and fungi exceeded 99.95% and 99.96%, respectively, indicating that almost all sequences were detected. Ace and Chao indices reflect microbial species richness, with higher values representing greater richness. Higher Shannon and lower Simpson values indicate higher community diversity (Gao, Peng, et al., 2024). For bacterial communities, DS-Y exhibited the highest richness based on Ace and Chao indices, followed by YN-K, DS-XF, LFQ, DS-DF and YN-T, with CS showing the lowest. Shannon and Simpson indices revealed high bacterial diversity in LFQ, DS-Y, YN-T, HQ-T, YN-K, and DS-XF. For fungal communities, Ace and Chao indices showed that DS-XF, YN-K and DS-DF had significantly higher richness than other samples, while CS was the lowest. DS-DF and DS-XF exhibited the highest Shannon and lowest Simpson indices, indicating superior fungal diversity.

Fig. 2.

Distributions of α-diversity estimators on OTU level. Ace index of bacteria (a) and fungi (b); Chao index of bacteria (c) and fungi (d); Shannon index of bacteria (e) and fungi (f). Significant differences are indicated by different letters (p < 0.05).

3.2.3. Microbial community structure analysis

A total of 167 genera, 94 families and 11 phyla of bacteria were identified by 16S rRNA gene sequencing. The major bacterial phyla were Firmicutes, Proteobacteria, unclassified_k__norank_d__Bacteria, and Actinobacteriota, with Firmicutes and Proteobacteria being the common dominant phyla (relative abundance >1%) across all samples (Fig. S3a). At the genus level, 18 main bacterial genera were detected, including Weissella, Leuconostoc, Lactococcus, Pediococcus, unclassified_k__norank_d__Bacteria, Enterobacter, Klebsiella, Acinetobacter, Rosenbergiella, Lactobacillus, Enterococcus, Streptococcus, Acetobacter, Staphylococcus, Exiguobacterium, unclassified_p__Proteobacteria, Bacillus and Serratia (Fig. 3a). The common dominant bacterial genera (relative abundance >1%) was Weissella. These above results are consistent with previous reports (Gao, Peng, et al., 2024; Peng et al., 2023; Zhang et al., 2022). DS-XF and YN-T exhibited the highest number of dominant genera, with 11 each. The composition of dominant bacterial genera varied significantly among sample groups. DS-DF, DS-XF and DS-Y were dominated by Weissella, Leuconostoc, Lactococcus, Streptococcus, Acetobacter and Staphylococcus. HQ-BT and HQ-T were dominated by Weissella, Lactococcus, Enterobacter, Acinetobacter and Exiguobacterium. The predominant bacterial genera of SJ-T and SJ-YJ were Weissella, Leuconostoc, Pediococcus and Enterococcus. Weissella, Leuconostoc, Lactococcus, Pediococcus, Enterobacter, Klebsiella, and Rosenbergiella were the principal bacterial genera characterizing YN-K and YN-T. LAB plays a crucial role in flavor generation and fermentation regulation of Huangjiu (Yang et al., 2020). LAB in FXXQ, including Weissella, Leuconostoc, Lactococcus, Pediococcus, and Lactobacillus, have strong acid-producing capacity, generating organic acids such as lactic acid and acetic acid through carbohydrate fermentation. Moreover, LAB enhance fermentation efficiency by regulating higher alcohols and reducing astringency and bitterness (Yang et al., 2024). Additionally, the high abundance of LAB in Hubei Huangjiu is strongly linked to increased organic acid levels, contributing to the distinct acidity characteristic of local Huangjiu (Peng et al., 2025).

Fig. 3.

Microbial community composition of bacteria (a) and fungi (b) at the genus level.

A total of 95 genera, 62 families, and 5 phyla of fungi were annotated by ITS gene sequence analysis. At the phylum level, the major fungi were Ascomycota, Basidiomycota, and Mucoromycota, with Ascomycota being the common dominant phylum (relative abundance >1%) across all samples (Fig. S3b). At the genus level, 15 main fungal genera were detected, including Saccharomycopsis, Wickerhamomyces, Candida, Kodamaea, Clavispora, Meyerozyma, Aspergillus, Apiotrichum, Rhizopus, Cyberlindnera, Dipodascaceae_gen_Incertae_sedis, unclassified_f__Dipodascaceae, Millerozyma, unclassified_o__Saccharomycetales and Lodderomyces (Fig. 3b). Except for YN-K, Saccharomycopsis was the common dominant genus across all samples, with relative abundances above 90% in CS, HQ-BT, HQ-T, JQ, LFQ, SJ-T, and SJ-YJ. The common dominant fungal genera of DS-DF and DS-XF included Saccharomycopsis, Wickerhamomyces, Candida, Clavispora, Meyerozyma, Aspergillus, Apiotrichum, Cyberlindnera and unclassified_f__Dipodascaceae, collectively accounting for 82.81% and 92.55% of total fungal community in DS-DF and DS-XF, respectively. The predominant fungal genera shared by HQ-BT, HQ-T and LFQ were Saccharomycopsis and Rhizopus. Notably, the relative abundance of Kodamaea in YN-K (38.60%) and Candida in YN-T (43.95%) were significantly higher than in other samples. These yeasts and filamentous fungi are vital for Huangjiu brewing, collectively driving saccharification, liquefaction, and alcohol fermentation (Yang et al., 2020). Specifically, yeast fermentation generates key flavor compounds, such as alcohols, esters, and volatile acids (Chen & Xu, 2010), but also secretes hydrolytic enzymes involved in the formation of their corresponding precursors. For instance, Saccharomycopsis secretes amylase, protease, and β-glucosidase to promote starch saccharification (Romano et al., 2003). Filamentous fungi such as Rhizopus and Aspergillus are commonly present in Xiaoqu and are integral to brewing (Yang et al., 2020). Rhizopus produces amylase and glucoamylase to degrade starch into glucose and synthesize flavor compounds like lactic acid and alcohols (Chen et al., 2020). Aspergillus secretes various amylases for saccharification and produces acid protease and carboxypeptidase, hydrolyzing proteins into nitrogen sources for yeast growth (Hayashi et al., 2021). Overall, the fungal community in Xiaoqu constitutes a functional core dominated by yeasts and filamentous fungi, whose synergistic and competitive interactions with bacteria determine the alcohol yield and flavor profile of Huangjiu.

Twenty-one bacterial species were shared across all samples (Fig. S4). DS-XF, DS-Y and YN-K exhibited the highest bacterial specificity with the largest number of unique species. In contrast, fungal composition showed more distinct differences, with only three fungal species shared by all samples. LFQ, DS-XF, YN-K, HQ-BT, and DS-DF had significantly more unique fungal species than others, reflecting their highly distinctive fungal community structures.

Compared with wheat Qu and Hongqu from other Huangjiu brewing systems, FXXQ exhibits distinct microbial characteristics. For fungi, wheat Qu is dominated by Aspergillus, Rhizopus, and Saccharomyces (Kang et al., 2022; Wang et al., 2021), and Hongqu is characterized by Aspergillus oryzae (Lv et al., 2013). FXXQ is distinguished by the dominance of Saccharomycopsis. For bacteria, FXXQ is dominated by Weissella, which differs from wheat Qu dominated by Bacillus (Tian et al., 2025; Xia et al., 2024) and Hongqu rich in Bacillus and Proteus (Huang et al., 2019). FXXQ also differs remarkably from typical Xiaoqu in southern China. For example, Hunan Xinhua Xiaoqu features Rhizopus and Mucor as dominant fungi with minimal Saccharomycopsis. Pediococcus, Enterobacter and Leuconostoc were the dominant bacteria, while Weissella was a minor genus (Xiao et al., 2022). In contrast, FXXQ lacked high-abundance Pediococcus and Enterobacter. The microbial uniqueness is reflected in the flavor differences between FXHJ and Huangjiu brewed with other types of Qu. The high-abundance LAB in FXXQ shapes its organic acid and ester aroma profiles, resulting in significantly higher total acid content in FXHJ compared with wheat Qu-fermented Shaoxing Huangjiu and establishing its regional sourness profile (Peng et al., 2025). Unlike the balanced multi-genus distribution in traditional Fujian Hongqu Huangjiu (Lv et al., 2013), FXXQ's dominance of Saccharomycopsis, with its efficient saccharification and alcohol synthesis capabilities, directly drives the characteristic flavor profile, featuring high β-phenylethanol and low isoamyl alcohol content (Gao, Peng, et al., 2024; Zhu et al., 2018). FXHJ also contains notable levels of ethyl lactate and 3-methyl-1-butanol (Liu et al., 2022; Zheng et al., 2025), which aligns with the profile of Xinhua Xiaoqu Huangjiu, whose core microbiota acts as the key driver for the synthesis of these flavor compounds (Xiao et al., 2022).

3.2.4. β-Diversity analysis of FXXQ microbial communities

To assess differences in microbial community structure among FXXQ, ANOSIM analysis was used to verify whether the between-group differences were significantly greater than within-group differences, thus evaluating the rationality of sample grouping. NMDS analysis based on Bray-Curtis distance was applied to analyze the degree of clustering and dispersion (Zhang et al., 2022). For both bacterial and fungal communities, between-group differences were significantly higher than within-group differences (Fig. 4a and c), supporting effective sample clustering. Adonis analysis based on the Bray-Curtis distance further quantified the contribution of grouping to community variation, revealing that grouping explained 88.35% of bacterial community variation (R² = 0.8835, P = 0.001) and 57.82% of fungal community variation (R² = 0.5782, P = 0.001) (Fig. 4b and d). The stress value of bacterial community was below 0.2, indicating reliable and interpretable separation. Eight samples were completely separated, yielding substantial differences in bacterial composition. This high variation may be attributed to the sensitivity of dominant bacterial genera such as Weissella and Leuconostoc to subtle differences in manufacturing processes, including moisture control and fermentation temperature (Peng et al., 2023). The fungal community had a stress value of less than 0.05, indicating excellent representation. DS-Y, DS-DF, DS-XF, YN-T and YN-K were completely distinguished, indicating their differences in fungal community composition. DS-DF, DS-XF, and DS-Y showed relatively unique fungal communities that may be associated with the specific Chinese herbal medicines added to their raw materials (Zhang et al., 2022). In contrast, fungal communities in other samples showed little variation and remained stable. This stability may be related to the dominance of Saccharomycopsis (relative abundance >90%), which plays a core functional role in starch saccharification and alcohol fermentation (Kang et al., 2022). These stable fungal communities were distinctly different from the five aforementioned samples, reflecting functional differentiation caused by raw material and process variations. Overall, fungal communities exhibited greater stability than bacterial communities (Liu et al., 2020).

Fig. 4.

Comparison of the intergroup differences in microbial communities of FXXQ. Box diagram of similarity analysis for bacteria (a) and fungi (c) based on Bray-Curtis distance algorithm; nonmetric multidimensional scaling (NMDS) analysis for bacteria (b) and fungi (d) at genus level.

3.2.5. Analysis of differences in microbial communities

To identify differential taxa across hierarchical levels among subgroups, LEfSe combined with linear discriminant analysis (LDA) was used for multilevel species difference analysis (Niu et al., 2025). A total of 255 differential bacterial taxa (LDA > 3.5, P < 0.05) were detected, including 5 phyla, 6 classes, 23 orders, 38 families, 70 genera, and 113 species (Fig. 5). Pediococcus and Nocardioides characterized subgroup CS. Leuconostoc and Streptococcus were the biomarkers for DS-DF. DS-XF was distinguished by Acetobacter and Staphylococcus. In DS-Y, Lactobacillus contributed most to group differences (LDA = 4.84, P < 0.05). HQ-BT was marked by Acinetobacter, while HQ-T was characterized by Lactococcus and Exiguobacterium. Fructobacillus and Enterococcus were the biomarkers for JQ and LFQ, respectively. In SJ-T, Enterobacter contributed most to the grouping differences (LDA = 4.74, P < 0.05). Weissella and Rosenbergiella were biomarkers for SJ-YJ and YN-K, respectively, with Klebsiella and Serratia distinguishing YN-T. Several LAB genera, including Pediococcus, Leuconostoc, Lactobacillus, and Weissella, were identified as biomarkers in different samples, consistent with the importance of LAB in Huangjiu fermentation (Yang et al., 2020; Yang et al., 2024). Compared with bacteria, fewer fungal groups exhibited significant differentiation, with no significantly enriched species in HQ-T, JQ or SJ-YJ (Fig. S5). Among the remaining samples, 151 differential taxa (LDA > 3.5, P < 0.05) were detected, covering 1 phylum, 8 classes, 13 orders, 23 families, 41 genera, and 65 species. Saccharomycopsis was the biomarker for CS. DS-DF was characterized by Meyerozyma and Millerozyma, while DS-XF was distinguished by Aspergillus and Apiotrichum. Clavispora and Candida were the biomarkers for DS-Y and YN-T, respectively. In YN-K, Wickerhamomyces was the primary biomarker (LDA = 5.45, P < 0.05), with Kodamaea and Hyphopichia also as significant biomarkers. Aspergillus and Saccharomycopsis, prominent in DS-XF and CS respectively, are associated with starch degradation and saccharification (Chen et al., 2020; Gao, Peng, et al., 2024; Zhang et al., 2022). The enrichment of Wickerhamomyces in YN-K aligns with elevated lipase activity (Shimizu et al., 2020). Notably, samples with high Saccharomycopsis abundance (HQ-T, JQ and SJ-YJ) lacked significant fungal biomarkers, while those with more balanced fungal communities (DS-DF, DS-XF and YN-K) exhibited multiple distinct biomarkers. This suggests that Saccharomycopsis dominance may suppress the detection of other fungal taxa in LEfSe analysis.

Fig. 5.

LEfSe analysis of differences in bacterial taxa of FXXQ. (a) Evolutionary distribution map; (b) diagram of LDA value distribution.

3.2.6. Functional microbial metabolic pathways

Microbial functions were predicted using the EggNOG database and categorized into 23 categories (Fig. 6a). While function S (function unknown; 19.45–21.39%) was the most abundant, function E (amino acid transport and metabolism; 7.50–11.30%) and G (carbohydrate transport and metabolism; 6.16–9.29%) were focused on due to their direct relevance to brewing. These categories correlated with the variations in saccharification power (98.60–317.00 U) and acid protease activity (9.64–330.97 U/g) across FXXQ samples. Specifically, DS-DF with higher saccharification power exhibited elevated fungal diversity with notable presence of Aspergillus, while CS with the highest acid protease activity showed higher abundances of Pediococcus, consistent with its proteolytic capacity (Jin et al., 2017). Metabolism-related functions dominated, accounting for 40.30% of all predicted functions, primarily within subcategories E (amino acid transport and metabolism; 24.29%), G (carbohydrate transport and metabolism; 19.88%) and P (inorganic ion transport and metabolism; 15.09%).

Fig. 6.

Functional microbial metabolic pathways analysis. (a) COG function classification abundance; (b) KEGG pathway classification abundance map.

According to the KEGG database, metabolism was the most abundant first-level pathway (75.35%). Carbohydrate metabolism (10.49%) and amino acid metabolism (6.29%) were the major secondary functions (Fig. 6b and Table S7). A total of 342 tertiary pathways were annotated, of which 118 pathways exhibiting relative abundances exceeding 0.1%. Key tertiary pathways included ko00500 (Starch and sucrose metabolism; 15.9%), ko00010 (Glycolysis/Gluconeogenesis; 12.6%) and ko00620 (Pyruvate metabolism; 11.1%). The pathway ko00500 supported the wide variation in saccharification power (98.60–317.00 U), while ko00010 and ko00620 constituted the metabolic cascade for ethanol synthesis, correlating with alcohol-producing capacity (7.80–13.74%vol). These core pathways are closely associated with VFCs in FXHJ, since fermentable sugars and ethanol serve as precursors for major alcohols and esters (Chen & Xu, 2010; Yang et al., 2017).

3.3. Analysis of VFCs in FXHJ

3.3.1. Characterization and quantification of VFCs

A total of 76 VFCs were identified by HS-SPME combined with GC–MS. These compounds included 21 esters, 20 alcohols, 11 acids, 6 ketones, 3 aldehydes, 4 sulfur compounds, 4 phenols, 4 lactones, and 3 others (2 terpenes and 1 acetal) (Table S8), with 30 VFCs shared by all samples. Notably, 10 VFCs were identified for the first time in FXHJ (Gao, Peng, et al., 2024; Liu et al., 2022; Tang & Peng, 2023; Yao et al., 2024; Zhang, Lin, et al., 2025; Zheng et al., 2025), including ethyl hydrogen succinate (10.38 μg/L), 4-penten-2-ol (2.80 μg/L), 3-methyl-2-butanol (41.20 μg/L), 1-hydroxy-2-propanone (13.07 μg/L), furaneol (2.68 μg/L), 1-phenyl-1,2-propanedione (13.08 μg/L), 5-ethyl-3-hydroxy-4-methyl-2(5H)-furanone (3.27 μg/L), 5-methyl-2-furancarboxaldehyde (41.90 μg/L), γ-decalactone (6.80 μg/L), and 1,1-diethoxyoctane (8.90 μg/L). These compounds have been reported in other fermented foods but not identified in FXHJ studies (Bai et al., 2025; Gao, Wang, et al., 2024; Geng et al., 2024; Piornos et al., 2020; Yao et al., 2021). Their aromatic characteristics (Table S9) and synergistic interactions with other VFCs collectively shape the flavor profile of FXHJ. The concentrations of various VFCs differed among samples (Fig. 7a). CS showed the highest total VFC content (37.52 mg/L), together with the highest concentrations of alcohols (24.15 mg/L), acids (4.78 mg/L), ketones (1.12 mg/L), and sulfur compounds (252.60 μg/L). HQ-BT ranked second in total VFCs abundance (35.17 mg/L) and the highest ester content (11.04 mg/L). In contrast, JQ had the lowest total VFC concentration (7.45 mg/L), with the lowest levels of esters (954.37 μg/L), alcohols (5.74 mg/L), acids (645.27 μg/L), ketones (58.01 μg/L), and phenols (24.13 μg/L). Moreover, DS-Y had the highest phenol content (561.41 μg/L), while SJ-T showed the highest concentrations of aldehydes (143.94 μg/L) and lactones (60.45 μg/L).

Fig. 7.

(a) Contents of volatile flavor compounds (VFCs) in FXHJ; (b) principal component analysis (PCA) of the VFCs in FXHJ; (c) variable importance for the projection (VIP) values of the VFCs in FXHJ (VIP > 1).

Esters and alcohols constituted the main VFCs in FXHJ, accounting for 7.30–31.41% and 54.17–85.46% of the total VFC concentration, respectively. Ester synthesis is mainly catalyzed by esterases secreted by microorganisms in FXXQ, with Leuconostoc, Staphylococcus, and Apiotrichum as the key esterase-producing genera that promote the esterification of alcohols and organic acids (Bai et al., 2025; Quan et al., 2023; Xiao et al., 2022), imparting fruity, floral, and sweet flavors to Huangjiu (Liu et al., 2020). Among detected esters, ethyl lactate was the most abundant (1.88 mg/L), consistent with previous research (Wang, Han, et al., 2020). Ethyl hexadecanoate, ethyl linoleate, ethyl myristate, ethyl oleate and diethyl succinate were also quantitatively significant, each with an average concentration exceeding 200 μg/L. Alcohol synthesis is primarily mediated by microorganisms such as Saccharomycopsis and Clavispora (Duan et al., 2024; Gao, Peng, et al., 2024). Phenethyl alcohol, contributing floral, sweet, fresh and bready with a rosey honey nuance (Luo et al., 2024), was the most abundant compound (13.11 mg/L). Moderate levels of higher alcohols such as 2,3-butanediol (423.77 μg/L) and 1-pentanol (246.06 μg/L) enhanced the flavor complexity of Huangjiu (Zhang et al., 2024). Acids regulate sweetness, astringency, and bitterness of Huangjiu, enhancing overall flavor harmony (Wang et al., 2019). Acetic acid was the most abundant acid (2.47 mg/L), accounting for 86.09–97.60% of total acids. It is mainly synthesized through carbohydrate fermentation by LAB such as Weissella and Lactobacillus (Peng et al., 2025). Notably, other compounds also contributed to the unique flavor profile of FXHJ. For instance, acetoin (267.69 μg/L) has been identified as an important flavor compound in Huangjiu (Zhang et al., 2024), characterized by a pleasant yogurt-like and creamy butter flavor. It serves as a precursor to bioactive compounds such as 2,3,5,6-tetramethylpyrazine (Xiao & Lu, 2014). Additionally, dimethyl trisulfide (20.26 μg/L) and 3-methylthiopropanol (70.52 μg/L) are typical aroma-active sulfur compounds in traditionally fermented alcoholic beverages, imparting cabbage-like odors and roasted flavors, respectively (Li et al., 2026). Furthermore, 2,4-di-tert-butylphenol (176.84 μg/L), furfural (8.30 μg/L), and γ-nonanolactone (51.10 μg/L), detected in this study, have also been reported in Shaoxing Huangjiu, Hakka Niangjiu, and Daixian Huangjiu (Bai et al., 2025; Wang, Han, et al., 2020). γ-Nonanolactone is synthesized via microbial bioconversion of oleic acid and α-linolenic acid (Bai et al., 2025), potentially involving fatty acid-metabolizing genera like Apiotrichum (Quan et al., 2023). Two terpenes named linalool (7.97 μg/L) and cadinene (36.43 μg/L) were identified, which may originate from traditional Chinese medicinal herbs (e.g., Polygonum hydropiper) added during the production of FXXQ (Zhang et al., 2022).

Principal component analysis (PCA) was used to assess similarities and differences in the composition of VFCs. The PCA plot revealed that the first principal component (PC1) explained 29.0% of the total variance, and the second principal component (PC2) explained 13.7% (Fig. 7b). All samples were within the 95% Hotelling T² confidence interval. DS-DF, DS-XF and DS-Y from the same manufacturer clustered closely, indicating high compositional similarity. Similarly, HQ-T, JQ, LFQ, YN-K and YN-T showed obvious similarity. However, HQ-BT and HQ-T differed significantly, potentially due to their different total sugar contents. Notably, CS was completely separated from other samples, indicating its unique flavor composition. Moreover, PLS-DA further clarified the diversity of VFCs among samples. Compounds with variable importance in projection (VIP) > 1 were regarded as key compounds for sample differentiation (Mahieu et al., 2023). Twenty-seven compounds such as decanoic acid (VIP = 1.385), 2-methyl-1-butanol (1.362), 4-methylvaleric acid (1.256), ethyl 2-hydroxy-3-methylbutanoate (1.256), ethyl 3-hydroxybutyrate (1.248), butyl acetate (1.244) and 2-Phenylethyl acetate (1.202) had VIP values higher than 1 (Fig. 7c). These substances were the main contributors to flavor differentiation among FXHJ samples.

3.3.2. Biosynthetic pathways of newly identified VFCs

To explore the formation of novel VFCs, their putative biosynthetic pathways were investigated. 3-Methyl-2-butanol and 4-penten-2-ol are produced via the Ehrlich pathway mediated by branched-chain amino acid aminotransferases and pyruvate decarboxylases (Hazelwood et al., 2008). γ-Decalactone is synthesized from ricinoleic acid through lipid β-oxidation (ko00061, ko00071) and catalyzed by extracellular lipases encoded by Lip2, Lip7, and Lip8 genes (Małajowicz et al., 2022). 5-Ethyl-3-hydroxy-4-methyl-2(5H)-furanone and furaneol are formed through the Maillard reaction (Blank & Fay, 1996). The former is linked to cysteine/methionine metabolism (ko00270) and arginine/proline metabolism (ko00330), while the latter is also associated with the pentose phosphate pathway (ko00030), regulated by 6GPL, 6PGDH, and TKT (Li et al., 2020). 5-Methyl-2-furancarboxaldehyde, another Maillard reaction product, accumulates through the pentose phosphate pathway (ko00030) and amino acid metabolism (Nor Qhairul Izzreen et al., 2016). 1-Phenyl-1,2-propanedione is an intermediate in phenylalanine metabolism (ko00360) formed by transamination and decarboxylation catalyzed by ARO10 (Vuralhan et al., 2003). Ethyl hydrogen succinate is associated with the TCA cycle (ko00020) and pyruvate metabolism (ko00620) through its precursor succinic acid (Orjuela et al., 2012). 1-Hydroxy-2-propanone is produced by microbial aldo-keto reductases and is related to glycolysis/gluconeogenesis (ko00010) (Vašková et al., 2025). However, the biosynthetic pathway of 1,1-diethoxyoctane during microbial fermentation remains unclear and requires further study.

3.4. Correlation analysis

3.4.1. Correlation analysis between physicochemical indicators and microbial communities of FXXQ

To investigate the relationships between physicochemical properties and major microbial communities in FXXQ, CCA and RDA were employed for bacterial and fungal genera (relative abundance >1%), respectively (Fig. 8a-b and Table S10). The choice of ordination method was based on the detrended correspondence analysis (DCA) results.

Fig. 8.

(a) Canonical correspondence analysis (CCA) between dominant bacteria genera and physicochemical parameters; (b) redundancy analysis (RDA) between dominant fungi genera and physicochemical parameters; (c) correlation heatmap of dominant microbial genera (abundance >1%) and volatile flavor compounds (contents >10 μg/L). Mo: moisture content; Ac: acidity; Sa: saccharification power; Pr: acid protease activity; Am: α-amylase activity; Li: lipase activity; Al: alcohol-producing capacity.

The bacterial community structure was significantly influenced by acid protease activity, moisture content, alcohol-producing capacity, saccharification power and lipase activity (P < 0.05). Pediococcus abundance was positively correlated with acid protease activity, due to its robust proteolytic capacity (Jin et al., 2017). Moisture content played a critical role in regulating microbial communities. Lactococcus and Acinetobacter showed significant positive correlations with moisture levels (Tian et al., 2025; Wen et al., 2021), and elevated moisture enriched Enterobacter populations (Li, Guan, et al., 2025). Consistent with previous findings, Weissella abundance was positively correlated with moisture (Zhao et al., 2023), alcohol-producing capacity (Zheng et al., 2023a) and α-amylase activity (Tian et al., 2025). Staphylococcus and Lactobacillus were positively correlated with acidity (Ma et al., 2024; Peng et al., 2023). In this study, Leuconostoc and Acetobacter also exhibited positive correlations with acidity, consistent with previous reports (Mitina et al., 2025; Wang, Pu, et al., 2024). In addition, Pediococcus was positively correlated with lipase activity, supporting its involvement in lipid metabolism (Abdi et al., 2024).

Alcohol-producing capacity exerted the most significant influence on fungal community structure (P = 0.001), followed by α-amylase activity (P = 0.002). Saccharomycopsis abundance was positively associated with α-amylase activity and alcohol-producing capacity, due to its efficient α-amylase secretion that degrades macromolecular substrates to release fermentable sugars and nutrients, promoting microbial alcohol biosynthesis (Wang, Yuan, et al., 2020; Zhu et al., 2023). Wickerhamomyces abundance was positively correlated with saccharification power and acidity (Peng et al., 2023; Tian et al., 2025). A similar positive trend was observed between saccharification power and Candida (Lin et al., 2022). Mixed fermentation with Starmerella bacillaris (syn. Candida zemplinina) significantly increased wine titratable acidity compared with Saccharomyces cerevisiae monocultures (Englezos et al., 2018), suggesting its role in acidification. Clavispora abundance showed a positive correlation with acid protease activity (Tian et al., 2025), and Clavispora lusitaniae PC3 has been documented as an efficient acid protease producer (Djekrif et al., 2024). Furthermore, acid protease activity and saccharification power showed synergistic relationships with Aspergillus abundance (Ma et al., 2024). Aspergillus-dominant starters can improve protease activity in fermentation starters (Liu et al., 2020), highlighting its role in protease production. Additionally, lipase activity was positively correlated with Wickerhamomyces and Candida abundance (Shimizu et al., 2020; Yalçın et al., 2014).

3.4.2. Correlation analysis between microorganisms and VFCs

Microorganisms play a significant role in regulating the flavor compounds of Huangjiu (Xu et al., 2024). Spearman correlation analysis was performed to evaluate relationships between 53 VFCs (average concentration > 10 μg/L) and microbial genera (relative abundance >1%) (Fig. 8c and Table S11). In the heatmap, darker colors indicate stronger correlations (red for positive, blue for negative), and squares with significance marks represent statistically significant correlations between corresponding VFCs and microbial genera. The analysis revealed that bacteria were primarily negatively correlated with volatile compounds, while most fungi were positively correlated with them.

As dominant bacteria, Leuconostoc and Staphylococcus play significant roles in substrate metabolism and flavor formation (Xiao et al., 2022), and both genera showed significant positive correlations with the esters pentyl butyrate, diethyl succinate, and ethyl oleate. For Leuconostoc, this correlation is attributed to its the metabolic capability in fatty acid biosynthesis and ester metabolism, encoding key enzymes for acyl donor generation, esterases for alcohol-acid esterification, and lipases for fatty acid decomposition to provide ester precursors (Li, Cui, et al., 2025). Streptococcus was positively correlated with most esters and acids, as well as with 4-ethylphenol. Similarly, Acetobacter exhibited positive correlations with esters and acids (Jiang et al., 2020; Qiu et al., 2021). For fungi, Saccharomycopsis, associated with alcohol synthesis enzymes (Gao, Peng, et al., 2024), showed significant positive correlations with most alcohols, especially 3-methyl-1-butanol. This compound is formed via the Ehrlich pathway, where leucine undergoes transamination, decarboxylation and reduction (Yan et al., 2022). The addition of Clavispora lusitaniae in Baijiu solid-state fermentation significantly increased the production of esters (e.g. ethyl hexadecanoate and 2-phenylethyl acetate) and phenethyl alcohol (Duan et al., 2024), which was confirmed in this study by a positive correlation between Clavispora and these compounds. Additionally, Apiotrichum showed highly significant positive correlations with pentyl butyrate, diethyl succinate, and ethyl myristate, consistent with its demonstrated capacity to synthesize ester precursors (Quan et al., 2023).

3.5. Practical implications for Xiaoqu production and quality control

This study provides direct practical guidance for the standardized production and quality control of FXXQ and FXHJ. Acid protease activity and alcohol-producing capacity of FXXQ are the key physicochemical factors regulating bacterial and fungal communities, respectively. In practical production, the microbial composition of Xiaoqu can be precisely regulated by adjusting koji preparation parameters. For example, optimizing raw material ratio, moisture content, and fermentation temperature would allow the selective enrichment of functional microorganisms such as Saccharomycopsis, Leuconostoc, and Staphylococcus, which are closely associated with high alcohol yield and ester formation. Previous studies have shown that optimizing the composition and inoculum size of multi-strain cocultures can markedly improve enzymatic activities (Yu et al., 2021). In addition to regulating temperature and moisture to ensure stable microbial activity, practical strategies for quality control include inoculating selected yeasts such as Saccharomyces cerevisiae to enhance aroma, and supplementing with specific herbs to inhibit undesirable bacteria and promote functional microbes (Zheng et al., 2023b; Zheng et al., 2025). Based on the microbial biomarkers identified by LEfSe analysis, a quality evaluation system for FXXQ can be established. In this system, Leuconostoc and Staphylococcus can be used as positive indicators of ester synthesis potential, and a relative abundance of Saccharomycopsis ≥ 90% can be adopted as a core indicator reflecting basic fermentation capacity of Xiaoqu.

3.6. Limitations and future perspectives

Although this study revealed correlations among physicochemical properties, microbial community structure of FXXQ, and flavor compounds of FXHJ, some limitations should be noted. First, laboratory fermentation differs from actual production environments, potentially leading to deviations in efficiency and types of flavor compounds. Second, key functional strains such as Saccharomycopsis and Leuconostoc were not isolated and verified by pure culture experiments, so their specific metabolic functions and contributions to flavor synthesis remain to be directly confirmed. Third, only volatile flavor compounds were analyzed, while important non-volatile flavor compounds such as free amino acids and polyphenols were not involved, which is not conducive to a comprehensive understanding of the flavor formation mechanism. Accordingly, several future research directions are proposed. To better reflect real production, in situ fermentation trials should be carried out to clarify the dynamic succession and interaction mechanism of physicochemical indices, microorganisms, and flavor metabolites. In addition, dominant strains should be isolated, purified, and functionally characterized to develop standardized composite starters. Moreover, multi-omics strategies (e.g., metagenomics combined with metatranscriptomics) are encouraged to reveal the metabolic pathways and regulatory mechanisms of key microorganisms at the molecular level. Finally, based on key physicochemical indicators and microbial markers screened in this study, a standardized production protocol and quality evaluation system for FXXQ can be established, which will provide technical support for the industrial production and quality improvement of FXHJ.

4. Conclusion

This study first employed an integrated multi-omics approach to systematically correlate the physicochemical properties and microbial community of FXXQ, as well as the flavor compounds in FXHJ. Multiple physicochemical indicators varied significantly across different samples. Weissella was identified as the common dominant bacterial genus in FXXQ, and Saccharomycopsis was the predominant fungal genus except in YN-K. Bacterial communities exhibited higher richness and diversity than fungal communities, whereas fungal community composition was more stable. LEfSe analysis identified 255 bacterial and 151 fungal taxa. A total of 76 VFCs were detected in FXHJ by HS-SPME combined with GC–MS, among which 10 compounds were reported for the first time in FXHJ. Phenethyl alcohol, acetic acid, and ethyl lactate were the most abundant compounds. Correlation analysis demonstrated that bacterial community structure was mainly affected by acid protease activity, moisture content, alcohol-producing capacity, saccharification power and lipase activity, while fungal community composition was predominantly influenced by alcohol-producing capacity and α-amylase activity. Leuconostoc and Staphylococcus showed significant positive correlations with multiple esters, and Saccharomycopsis was positively associated with most alcohols. These findings address the research gaps of FXXQ and provide theoretical support for optimizing the traditional FXHJ brewing process and improving product quality.

CRediT authorship contribution statement

Xuanjing Ren: Writing – review & editing, Writing – original draft, Software, Methodology, Investigation, Data curation, Conceptualization. Qingya Wen: Writing – review & editing, Methodology, Investigation. Lulu Cui: Validation, Resources, Project administration. Tianyi Cao: Supervision. Die Hu: Validation, Resources, Project administration. Hongqin Liu: Supervision. Jihong Wu: Supervision. Qing Ren: Writing – review & editing, Supervision, Methodology, Investigation, Funding acquisition. Mingquan Huang: Writing – review & editing, Supervision, Methodology, Investigation, Funding acquisition. Dongrui Zhao: Supervision. Hehe Li: Supervision.

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

Data availability

Data will be made available on request.