Co-fermentation of Pichia kudriavzevii and Millerozyma farinosa modulates fungal community and enhances flavor in Yangjiang douchi

Abstract

Two aromatic yeast strains, YPKJ (Pichia kudriavzevii) and YMFJ (Millerozyma farinosa), were isolated and identified from Yangjiang fermented douchi. The optimal fermentation conditions—9% salt content, 34 °C, 4% inoculum size, and a 1:2 inoculum ratio—were determined based on amino acid nitrogen and total acidity. Using a Co-fermentation strategy combined with MiSeq sequencing and gas chromatography–mass spectrometry (GC–MS), we investigated the influence of microbial communities and flavor compounds during douchi post-fermentation. Results revealed 156 volatile compounds identified by GC–MS, with their content increasing approximately 6.7-fold during ZF fermentation. Specifically, the ester content in Z10 was 1.38 times higher than that in Y30, and the phenethyl alcohol content in Z10 was 2.4 times higher than that in Y20. This study provides a theoretical and technical basis for regulating the microbial community in fermented black bean production and for optimizing industrial manufacturing processes.

Graphical abstract

Highlights

• •Aromatic yeasts Pichia kudriavzevii and Millerozyma farinosa were isolated from Yangjiang douchi.
• •Co-fermentation increased succinic acid, GMP, and umami amino acids (Asp and Glu).
• •Co-fermentation altered fungal community succession and elevated the abundances of the inoculated yeasts.
• •Co-fermentation reduced key off-odors while enhancing fruity, floral, and smoky aromas.

1. Introduction

Douchi is a typical representative of traditional Chinese fermented soybean products and renowned for its distinctive umami taste and rich aroma, derives its umami and aroma from microbial metabolism during fermentation (Zhou et al., 2024). Based on the different types of microorganisms used in the koji-making stage, are divided into Aspergillus-type, Rhizopus-type, Bacterial-type, and Mucor-type douchi (He et al., 2019; Jang, Oh, Lim, Kim, & Kim, 2021). Moreover, Yangjiang fermented douchi are a mould-type fermented bean paste made from douchi, with a salt content exceeding 10% during production (Chen et al., 2021; Wu et al., 2024). Both controllable and uncontrollable factors in fermentation alter the microbial community of the paste, thereby influencing its quality and flavor. Notable genera include Corynebacterium_1, Pseudonocardia, and Beauveria (Chen et al., 2022). Research using high-throughput sequencing has identified the dominant microorganisms in this bean paste as primarily Aspergillus species, Saccharomyces cerevisiae, and Bacillus species. In Aspergillus-type douchi production, critical flavor-forming volatiles develop predominantly during the post-fermentation stage (Jiang et al., 2026). Nevertheless, due to instability in flavor formation during post-fermentation and high-salt inhibition of beneficial microorganisms particularly yeasts, limitations exist in single-strain applications.

During douchi production, yeasts are key functional microbiota that metabolize soybean carbohydrates via the glycolytic pathway to produce ethanol and intermediate precursors, thereby establishing the biochemical foundation for the biosynthesis of characteristic aroma compounds such as esters and alcohols (Chen et al., 2023). Current research predominantly focuses on screening the protease-producing strains and their applications in koji-making, while largely overlooking the critical role of key functional microbiota that driveflavor formation during the post-fermentation phase. Guo, Hou, Wang, and Wang (2025) isolated and identified four enzyme-producing strains from Yongchuan douchi utilizing traditional culture methods and high-throughput sequencing techniques, namely Aspergillus oryzae, Mucor circinelloides, Wickerhamomyces anomalus, and Penicillium griseofulvum. The complex strain of Hansenula debaryi yeast, isolated from Korean fermented soybean products, exhibits rich flavor characteristics and produces diverse volatile flavor compounds, including esters and alcohols. It also possesses high bioconversion activity that converts ferulic acid into 4-vinylguaiacol—a characteristic soybean flavor compound—thereby enhancing the flavor and quality of fermented products (Jeong et al., 2022).Therefore, screening flavor-regulating microbial strains from douchi is crucial for enhancing and improving product flavor profiles. Zhao et al. (2023) utilized high-throughput sequencing to investigate the microbial structure of Aspergillus-type douchi, which revealed that during the post-fermentation stage, the dominant microbiota associated with flavor formation primarily included Millerozyma, Pichia, and Candida spp., among others. This study provides important references for the screening and functional characterization of aroma-producing yeasts in douchi production.

While pure culture fermentation offers microbial controllability, using single strains often results in limited flavor complexity. In contrast, multi-strain Co-fermentation enhances flavor diversity through metabolic complementarity and synergistic enzymatic systems. Biscudii yeast synthesizes lipids and utilizes ethanol and acetic acid via esterases to produce aromatic compounds, chiefly ethyl acetate (Hua et al., 2025). Conversely, Miller yeast enhances soy sauce's volatile ester content mainly by synthesizing esters like ethyl acetate and ethyl isovalerate through esterification and alcoholysis (Jiang et al., 2021). Yeast could produce aroma substances such as esters, alcohols, and phenols through metabolism, which the synergistic fermentation of multiple yeast species helps to enhance the flavor (Wang, Xie, Feng, Huang, & Zhao, 2024). Previous studies have confirmed that a single yeast strain in fermented foods could only produce simple aromatic components, whereas mixed microbial cultures (such as P. kudriavzevii and Saccharomyces cerevisiae) significantly enhance the diversity of ester compounds through synergistic metabolism (Devi, Anu-Appaiah, & Lin, 2022; Li et al., 2025; Li et al., 2025). A key limitation in current research lies in the lack of functional validation, particularly regarding Co-fermentation synergies of yeasts (e.g., Millerozyma, Pichia) identified via high-throughput sequencing in high-salt douchi systems, along with insufficient investigation into their active modulation of microbial community structure, comprehensive impact on global flavor profiles.

Therefore, this study aims to screen and identify yeast strains that possess the traditional aroma of douchi. Building on this, we further focus on multi-strain synergistic effects, investigating the comprehensive impacts of Co-fermentation with different yeast species on microbial community dynamics, organic acid accumulation, generation of free amino acids and flavor-enhancing nucleotides, as well as synthesis of volatile flavor compounds during douchi fermentation. This study integrates dynamic tracking of flavor precursors and key metabolites, not only providing data support for deciphering the synergistic aroma-producing mechanisms of mixed microbial communities, but also offering process optimization pathways for the industrial production of traditional douchi. These findings contribute to achieving precise regulation of fermentation processes and stable enhancement of product quality, thereby facilitating the modernization of traditional fermented foods.

2. Materials and methods

2.1. Douchi sample collection

Previous studies have indicated that yeast primarily exists during the post-fermentation stage (Jiang et al., 2026), and the production process is shown in Fig. 1. Therefore, Yangfan douchi (YF) samples (Yangfan Food Co., Ltd., Guangdong, China) from the salt-mixing stage (before the start of post-fermentation), and 10, 20, and 30 d of post-fermentation were collected as raw materials for yeast screening and as control group for fermentation. Additionally, douchi samples from the 5 d of koji-making were collected for subsequent post-process optimization, designated Y0, Y10, Y20, Y30, and F5, respectively.

Fig. 1.

The fermentation production process of Aspergillus-type douchi.

2.2. Yeast isolation

Referring to the method of yeast isolation and purification described by (Lai et al., 2022), 10 g of douchi was mixed with 90 mL sterile saline (containing glass beads) to prepare a 10⁻¹ bacterial suspension, enriched at 30 °C (150 × rpm, 24 h), and serially diluted to 10⁻² - 10⁻⁶. Aliquots (200 μL) of 10⁻⁴, 10⁻⁵, and 10⁻⁶ dilutions were spread onto Yeast extract powder peptone glucose (YPD) or Bengal Red agar plates (Hope Bio-Technology Co., Ltd. Qingdao, China) and incubated at 30 °C for 48 h. Morphologically distinct colonies were purified 3 to 4 times using a sterile loop, transferred to YPD or Bengal Red slant media, cultured at 30 °C for 48 h, and stored at 4 °C for subsequent use.

2.3. Initial screening of aroma-producing yeast based on sensory evaluation and secondary screening by total ester determination

To evaluate the aroma-enhancing performance of the yeast, a 5 μL disposable sterile inoculation ring was used to attach a ring of yeast strain to the YPD liquid medium and activated for 48 h at 30 °C and 150 × rpm. The activated yeast strains were inoculated into the new YPD liquid culture medium at a 1% (v/v) inoculation rate, and the YPD liquid culture medium of uninoculated yeast was used as a blank control, and cultured in a biochemical incubator at 30 °C at 150 × rpm for 48 h. The aroma-enhancing ability of the yeast was preliminarily evaluated by the sensory evaluation categorizing the scents into six types: mellow, floral, fruity, sweet, stink, and ammoniac odors (Tang et al., 2020). The results were evaluated on a scale of 0 to 10 based on the intensity of the aroma, with higher scores indicating stronger aroma, and the strains exhibiting displeasure odors, such as stink or ammoniac odors production were discarded (; Li, Zhao, et al., 2025). The strains obtained from the preliminary screening were inoculated into a wort liquid medium and fermented at 30 °C for 48 h, and the total ester content of the yeast fermentation broth was then determined following the method described by Zhang et al. (2017). The strains with relatively high total ester content, as determined by secondary screening, were preserved in glycerol and stored at −80 °C.

2.4. Yeast identification

Single colonies were selected to make microscope smears, which were stained with 0.1% Lv's alkaline Meilan staining solution (Hope Bio-Technology Co., Ltd. Qingdao, China). The basic characteristics of the strain were determined after staining, and the morphological characteristics of the cells were observed by biological microscopy. The DNA of the yeast genome was isolated and extracted by a Magen magnetic bead DNA extraction kit (Magen Ltd., Guangzhou, China). The purity, concentration, and integrity of nucleic acids were detected by NanoDrop 8000 (Thermo Fisher Scientific, Waltham, MA, USA) and 1% agarose gel electrophoresis (Sangon Ltd., Shanghai, China), and 26S rDNA universal primers NL1 and NL4 were used amplified the D1/D2 region gene sequence of the strain. All reactions were performed in 30 μL of a hybrid system containing approximately 15 ng of genomic DNA, 15 μL 2 × Taq PCR Master Mix (GeneTech Inc., Shanghai, China), Upstream primer, and downstream primer 2.0 μL, respectively, and ddH₂O used to adjust the volume. According to the PCR amplification conditions provided by (Fazio et al., 2024), the amplification was performed in the Veriti 96 well thermal cycler (Applied Biosystems, Inc., CA, USA). The quality of PCR products was detected by 1% agarose gel electrophoresis, and the specificity of the amplified products of each sample was determined by the electrophoresis strip of PCR products. Additionally, PCR products matching the size of the target fragment were Sanger sequenced on the 3730xL DNA Analyzer (Applied Biosystems, Inc., CA, USA). The homology (similarity ≥99%) of the sequenced sequences was compared and analyzed in the NCBI database, and the phylogenetic tree was constructed using the neighbor-joining method in MEGA 11. software.

2.5. Determination of characteristics and growth curve of yeast

The activated perfuming yeast strains were inoculated into 100 mL YPD liquid medium with different NaCl concentrations according to 1% (v/v) inoculation rate, and the concentration gradient of NaCl added to the medium was 0%, 7.5%, 10%, 12.5%, 15%, and 17.5%. The culture oscillated at 150 × rpm at 30 °C for 48 h, and the OD₆₀₀ nm value of bacteria suspension was measured by UV spectrophotometer (Shimadzu, Tokyo, Japan) with YPD liquid medium without bacteria as blank control. Meanwhile, the activated aromatic yeast strains were inoculated in 100 mL YPD liquid medium at 1% (v/v) and oscillated at 150 × rpm at 30 °C for 72 h. Additionally, samples were taken every 3 h during the culture period to measure the OD₆₀₀ nm value of bacterial suspension, and the test was repeated 3 times. Activated aromatic yeast strains were inoculated onto Blood agar plates (Hope Bio-Technology Co., Ltd. Qingdao, China) and incubated inverted at 30 °C for 48 h. The presence of hemolytic zones around the colonies was observed, with the evaluation criteria referred to the methodology described in (Shang et al., 2019).

2.6. Preparation of the fermentation starter

Referring to the research method of (Gao et al., 2025) with slight modifications, YPKJ and YMFJ were subcultured twice in YPD broth and incubated at 30 °C for 24 h. The cultures were then centrifuged at 4 °C and 8000 × rpm for 5 min using a refrigerated high-speed centrifuge, washed twice with 0.9% sterile saline (w/v), and adjusted to an initial yeast concentration of 10⁶ cfu/g.

2.7. Optimization of the post-fermentation process of douchi and co-fermentation fermentation

Based on the original fermentation process, the effects of salt addition (8%, 9%, 10%, 11%, and 12%), fermentation temperature (28 °C, 30 °C, 32 °C, 34 °C, and 36 °C), inoculation amount (2%, 3%, 4%, 5%, 6%), inoculation ratio (1:1, 1:2, 1:3, 2:1, and 3:1) on amino acid nitrogen and total acid content were investigated. Each experimental group was conducted in triplicate.

The optimal post-fermentation process conditions obtained through a single factor were used for Co-fermentation fermentation, which is the fermentation process in Fig. 1. After washing the koji beans that had been made for 5 days, the moisture content was controlled to 50%–60%, salt was added, and starter cultures (dual-strain co-inoculation) were inoculated for 30 days of piling and fermentation (three parallel groups). The finished product was named ZF (inoculated group), which was taken and analyzed on the 10th, 20th, and 30th days of post-fermentation (Z10, Z20, Z30), and after salt-mixed inoculation before fermentation (Z0).

2.8. High-throughput sequencing of microbial flora composition

Douchi microbial DNA was extracted using the E.ZN.A Kit (Omega Bio-tek, Norcross, GA, USA) and its quality were assessed via Nanodrop 8000 (Thermo Fisher Scientific, Waltham, MA, USA), Qubit™ 4.0 (Thermo Fisher Scientific, Waltham, MA, USA), and 2% agarose gel electrophoresis (Sangon Ltd., Shanghai, China) before storage at −80 °C. The fungal ITS region was amplified with primers ITS1/ITS2. Following Qiu et al.'s protocol (Qiu et al., 2018), PCR amplification was performed in a 30 μL reaction system containing 15 μL 2 × Hieff® PCR Master Mix (Yeasen, Shanghai, China), 10–20 ng of the template DNA, 1 μL of each primer, and ddH₂O to adjust the volume, using a T100TM thermal cycler (Bio-Rad, San Diego, CA, USA). The second-round PCR adjusted the template DNA to 25 ng and reduced cycles to 5. Purified PCR products with magnetic beads were sequenced on the Illumina MiSeq PE300 platform after quantification with Qubit™ 4.0 to determine sequencing ratios.

2.9. Physicochemical analysis of douchi

The content of reducing sugars in douchi was determined using the 3,5-Dinitrosalicylic acid (DNS) method, as described by Yuan et al. (2023) with minor modifications. Take 1 g of the douchi sample mix it with 2.0 mL of DNS reagent, and place it in a boiling water bath for 10 min. After cooling, the solution was made up to 25 mL with ultrapure water. The absorbance was measured at 540 nm using an ultraviolet spectrophotometer (Purkinje General Instrument Co., Ltd., Beijing, China) and a glucose standard curve. The methods for determining salt content, total acid, and amino acid nitrogen are consistent with those used in previous studies (Jiang et al., 2026).

2.10. Sensory analysis

A sensory evaluation panel was established following a two-week systematic training program, during which members were rigorously selected and professionally trained to describe the color, texture, flavor, and appearance of fermented black bean samples. The panel comprised ten members (aged 22–28, five male and five female) from Guangdong Ocean University. The sensory assessment was conducted in a well-ventilated food laboratory that provided ample lighting and space. The evaluation system encompassed four attributes: color (intensity, uniformity, glossiness); texture (saltiness, hardness, chewiness); flavor (umami, bitterness, acidity, fermented bean aroma, ammoniacal odor, beaniness, mouldiness); and appearance (granularity, shape integrity). Independent ratings for these attributes employed a 1–5 scale (5 = very much liked, 1 = disliked), with higher scores denoting superior sensory quality (Jiang et al., 2021).

2.11. Identification of aroma-producing VOCs in yeast and douchi by GC–MS

The Volatile Flavor Compounds (VOCs) of yeast-fermented broth or fermented douchi were collected by headspace solid phase microextraction (HS-SPME). Samples of 5 mL yeast-fermented broth or 5 g douchi samples were taken and added with 0.5 g NaCl and 5 μL of 2-methyl-3-heptanone standard solution (81.6 μg/mL) as the internal standard, and the mixture was sealed in a 20 mL headspace bottle. The samples were extracted by CAR-DVB-PDMS needle (50/30 μm, Supelco) and HS-SPME for 40 min. The extraction needle was then inserted into the Model 8890 GC–MS (Agilent, Waltham, MA, USA) injector and resolved at 250 °C for 5 min.

Used previously published procedures with slight modifications (Zhao et al., 2024). DB-WAX column (60 m × 0.25 mm × 0.25 μm; Agilent, Waltham, MA, USA), selected helium (purity >99.99%) as carrier gas and the flow rate was 1.0 mL/min, and non-diversion mode. The initial column temperature was set at 40 °C, then increased to 120 °C at 4 °C /min and held for 4 min, increased to 180 °C at 2 °C/min and held for 5 min, and finally heated to 220 °C at 4 °C/min and held for 3 min. VOCs were matched and identified by comparing their mass spectra with standard compound data in the NIST20 mass spectrometry database (Gaithersburg, MD, USA). The relative content of VOCs was calculated based on the ratio of the peak area of VOCs in the test sample to the peak area of the internal target. The calculation of Odor Activity Value (OAV) was based on the ratio of the relative content of VOCs to the threshold value, which refers to the threshold information for water in Van Gemert's version of the Compound Threshold Compilation (Van Gemert, 2018).

2.12. Determination of organic acids in fermented douchi

According to Chinese standard GB 5009.157–2016, the determination was performed by high-performance liquid chromatography with Agilent Poroshell 120, EC-C18 column (4.6 mm × 250 mm, 5 μm). The mobile phase was methanol: 0.1% phosphoric acid, the ratio was 12:88 (v/v), the column temperature was 30 °C, the flow rate was 0.5 mL/min, and the injection volume was 20 μL. A standard working solution of organic acids (succinic acid, lactic acid, malic acid, and citric acid) at concentrations of 2500 μg/mL, 1000 μg/mL, 500 μg/mL, 250 and 125 μg/mL, respectively) was accurately prepared. The samples were filtered by a 0.22 μm organic filter membrane (Jinteng Experimental Equipment Co., Ltd., Tianjin, China), and the peak area was used to quantify organic acids.

2.13. Determination of flavor nucleotides

The content of flavor nucleotides in douchi was determined using the method described by Liu et al. (2024). An HPLC equipped with a Poroshell 120 EC-C18 column (100 mm × 4.6 mm, 4 μm) was used, which with UV detection at a wavelength of 254 nm. The mobile phase consisted of methanol and 0.01% potassium dihydrogen phosphate in a ratio of 2:98 (v/v), with a flow rate of 0.5 mL/min, the column temperature was 30 °C, and the injection volume was 10 μL. Weigh 10 mg each of adenosine monophosphate (AMP), guanosine monophosphate (GMP), cytidine monophosphate (CMP), uridine monophosphate (UMP), and inosine monophosphate (IMP) standards, and dilute them with ultrapure water to prepare mixed standard solutions with concentrations of 5, 10, 25, 50, and 100 μg/mL. The sample was dissolved in ultrapure water and filtered through a 0.45 μm membrane. The flavor nucleotides in douchi were quantified using the standard curve derived from the peak areas of the mixed standards.

2.14. Determination of free amino acids (FAAs) in fermented douchi

Amino acid analyzer L-8900 (Hitachi Ltd., Tokyo, Japan) was used for atomic absorption spectrometry analysis of the sample, and the Chinese national standard GB5009.124–2016 was adopted, with slight modifications. 2 g of fermented douchi were mixed with 1 mL of 5-sulfosalicylic acid (Macklin Biochemical Technology Co., Ltd., Shanghai, China) and left for 55 min at 4 °C. Centrifuge at 12000 ×g for 10 min, filter with 0.22 μm organic filter membrane, and the injection volume was 50 μL. The preparation method of the mobile phase was consistent with that of Chen et al. (2024).

2.15. Statistical analysis

All experiments were repeated three times, and the data were expressed as mean ± standard deviation. Origin 2021pro software was used to draw pictures. OPLS-DA analysis was conducted using SIMCA software version 14.1. Besides, MEGA11. software was used to draw the yeast phylogenetic tree, and SPSS 22.0 software was used for statistical analysis. The raw sequencing data for the Millerozyma farinosa and Pichia kudriavzevii strains have been deposited in the National Center for Biotechnology Information (NCBI)in the United States.The accession numbers are PX466804 and PX466378, corresponding to the germplasm numbers YMFJ and YPKJ, respectively. Furthermore, the use of the OmicStudio cloud tool to create microbial communities associated with key flavor substances of network diagram (https://www.omicstudio.cn/tool).

3. Results and discussion

3.1. Screening and identification of aromatic yeast

3.1.1. Analysis of primary and secondary screening results

Fig. 2A shows that 5 yeast strains were isolated from douchi and designated YPKJ, YMFJ, YMF3, YMF4, and YMF5. Among these, YMF3 and YMF5 primarily exhibited off-flavors characterized by stink and ammoniac odors, whereas YPKJ, YMFJ, and YMF4 produced fruity, sweet, and alcoholic aromas. Consequently, these three strains (YPKJ, YMFJ, and YMF4) were selected for secondary screening.

Fig. 2.

Results of Screening and Identification of Aroma-Producing Yeast Strains. Sensory evaluation score and total ester content of yeast, (A) Evaluation score; (B) Total ester content; Yeast colony morphology and microscopic morphological characteristics, the colony morphology (C) YPKJ and (D) YMFJ, (p < 0.05); the colony microscopic morphological (E) YPKJ and (F)YMFJ;

The amplification results and phylogenetic tree of the PCR products of YPKJ and YMFJ, (G) electrophoresis diagram; (H) phylogenetic tree.

Esters are the main aroma-producing substances in yeast fermentation products， which primarily generate fruity, honey, and floral aromas. These ester compounds represent a crucial parameter for assessing the fragrance-generating potential of yeast strains (Tang et al., 2020). The three yeast strains isolated during preliminary screening were subjected to total ester content analysis employing the saponification reflux technique. With total ester contents of 1.62 ± 0.15 g/L (YPKJ) and 1.28 ± 0.04 g/L (YMFJ), these two strains demonstrated significantly higher production levels compared with others, as illustrated in Fig. 2B. Therefore, these two strains can be determined to possess excellent ester-producing capacity. Furthermore, Li et al. (2016) isolated a strain of Saccharomyces rouxii from fermented chili sauce, and the total ester content of this strain was 0.55 g/L. Therefore, it could be concluded that these two strains have relatively good ester-producing capabilities.

3.1.2. Morphological characterization and molecular biological identification analysis

Through enrichment, isolation, and purification processes, followed by preliminary and secondary screening, two yeast strains with strong aroma-producing capabilities were obtained from the 10th and 30th days of post-fermentation, designated as YPKJ and YMFJ, respectively. The colony analysis showed that both YPKJ and YMFJ presented typical yeast-shaped ellipses (Fig. 2C-D). Additionally, the ML31 biological microscope (Mshot Technology Co., Ltd., Guangzhou, China) also confirmed the elliptical shape. YPKJ exhibited budding reproduction (Fig. 2E-F).

To further determine the taxonomic status of strains YPKJ and YMFJ, their 26S rDNA was amplified by PCR. Electrophoresis showed a clear and distinct band of the amplified fragment at approximately 600 bp, in line with the target size (Fig. 2G). Additionally, Sanger sequencing data were submitted to NCBI, followed by BLAST homology analysis and construction of a phylogenetic tree. The results showed that YPKJ exhibited 100% sequence similarity to Pichia kudriavzevii, and YMFJ showed 100% similarity to Millerozyma farinosa (Fig. 2H).

3.1.3. Analysis of strain characteristics and growth curve

To explore the salt-tolerant properties of YPKJ and YMFJ, their cell activities under different salt-content environments were measured, as shown in Fig. S1-A. The study indicates that both yeast strains exhibit inhibited growth at a 7.5% salt concentration but remain viable at 17.5%, suggesting their potential suitability for high-salt fermentation in douchi production. The growth rate of YPKJ and YMFJ was explored by measuring the light absorption value at OD₆₀₀ nm, and the results are shown in the Fig. S1-B. The results showed that the exponential growth periods of YPKJ and YMFJ were 0–45 h, and the fastest growth periods of YPKJ and YMFJ were 0–9 h and 0–21 h, respectively. Additionally, the growth rate of YPKJ slowed down and entered the plateau period at 48–66 h, while the plateau period of YMFJ was 3 h longer than that of YPKJ. Thus, the results indicate that both strains have good growth rates. It is noteworthy that both strains did not exhibit hemolytic activity (Fig. S1-C), preliminarily demonstrating their safety profile. In summary, these two strains are considered to have potential for application in high-salt fermentation environments.

3.2. Analysis of VOCs in fermentation broth

Detection of VOCs in yeast fermentation broth can reflect the aroma-producing characteristics of yeast strains. The HS-SPME-GC–MS analysis revealed that the fermentation broth of YPKJ and YMFJ contained 63 and 59 flavor substances, respectively, with a total content of 10,193.4 μg/kg and 3874.66 μg/kg, which was 13.88 times and 5.27 times higher than that of the control group (YPD) (Fig. 3A and Table S1). Both strains significantly increased the types of esters, alcohols, and ketones, and the ester types of YPKJ reached 16 (YPD had only 1 type). However, the number of alcohols, ketones, and esters in YMFJ was 13, 5, and 4 more than that of YPD, respectively. Esters are key products of microbial metabolic activity, with their formation closely linked to enzymatic catalysis and environmental adaptation. This study shows that the total content of esters in YPKJ and YMFJ is (2685.93 μg/kg) and (21.07 μg/kg) respectively, which is significantly higher than that of the YPD control group (Table S1). Previous studies have confirmed that Pichia kudriavzevii and Millerozyma farinosa could promote ester synthesis (Jiang et al., 2021; Pereira et al., 2017). Furthermore, microorganisms could convert sugars or other substrates into alcohols through fermentative metabolic pathways. Compared to YPD (386.80 μg/kg), the total amount of alcohols in the fermentation liquids of YPKJ (5199.37 μg/kg) and YMFJ (1916.92 μg/kg) has significantly increased (Table S1). Among them, phenethyl alcohol was a unique higher alcohol specific to YPKJ and YMFJ, which higher alcohols may be associated with the metabolism of branched-chain acids and yeast strains (Chen et al., 2023; Pereira et al., 2017)_. Moreover, ketones, including 3-octanone and 2-undecanone, were predominantly detected in YMFJ. Therefore, it could be inferred that YPKJ and YMFJ are conducive to the formation of esters, alcohols and ketones.

Fig. 3.

Analysis of volatile flavor compounds in the yeast fermentation broth. (A) Venn diagram of species; (B) OPLS-DA differential analysis plot; (C) Cross-validation plot by 200 permutation tests; (D) Clustered heatmap of key flavor substances.

Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) was performed to analyze differences in VOCs between yeast-fermented and non-inoculated fermentation broths. As shown in Fig. 3B, the OPLS-DA model exhibited strong interpretability and predictive performance, with fitting parameters R²Y = 0.993 and Q² = 0.984. YPD (the second quadrant), YPKJ (the third and fourth quadrants), and YMFJ (the first quadrant) were distributed in distinct quadrants, indicating that there are significant differences in volatile flavor compounds among them. Meanwhile, after 200 permutation tests, the OPLS-DA model was verified to be reliable (Fig. 3C). Spearman's correlation clustering heatmap (Fig. 3D) shows the association between 13 key flavor compounds (OAV ≥ 1) (Table S2) and yeast strains. The YPKJ strain primarily produced fruity (banana) and floral (rose) aromas, in which key compounds (ethyl isovalerate, ethyl acetate, ethyl isobutyrate, phenethyl alcohol etc.) showed significantly positive correlations with the strain. The YMFJ strain was characterized primarily by alcohols and ketones (2-undecanone, and 2-nonanol), contributing creamy, fruity, and floral aromas. YPD observed a positive correlation with phenylacetaldehyde, which might be attributed to the degradation of saccharides in the culture medium.

3.3. Analysis of single-factor optimization results for the post-fermentation process

As shown in Fig. 4, the amino acid nitrogen (1.27%) and total acid (1.77%) reached peak values at 9% salt addition. At 34 °C fermentation temperature, their highest levels were 1.02% and 2.07%, respectively. The optimal 4% inoculum size yielded 1.15% (amino acid nitrogen) and 1.67% (total acid), while a 1:2 inoculation ratio achieved maximum values of 1.13% and 2.45%. Thus, the optimal fermentation conditions were 9% salt content, 34 °C, 4% inoculum size, and 1:2 inoculation ratio.

Fig. 4.

Effects of different fermentation conditions on amino acid nitrogen and total acid content in the douchi. (A) Salt content; (B) Temperature; (C) Inoculation amount; (D) Inoculation ratio.

3.4. Fungal community succession during douchi fermentation

3.4.1. Microbial diversity analysis of different types of fermented douchi

Alpha diversity analysis reflects the richness and diversity of microbial communities within each sample group. The richness of microbial communities is typically assessed using the Ace and Chao indices (Yang, Yang, Tu, & Wang, 2016). The Simpson and Shannon indices are well-established metrics for estimating microbial diversity within samples. Notably, the Simpson index decreases with increasing species richness, while the Shannon index exhibits the opposite trend (Cai et al., 2021). Analysis of Fig. 5 (A-B) revealed significant differences in Ace and Shannon indices between ZF and YF during post-fermentation, which exhibited a characteristic decrease-to-increase pattern from Z0 to Z30, showed an inverse trend to YF in indices, and maintained relatively lower values for both diversity and richness measures. Previous studies have shown that adding yeast (Saccharomyces cerevisiae) to Baijiu results in a trend of initially decreasing and then increasing fungal abundance in the Baijiu, and also leads to a decrease in the diversity and richness of bacteria and fungi (Qiu et al., 2024). This could be due to the fact that the suppression of indigenous microbiota (such as Aspergillus spp.) by co-inoculated strains YPKJ and YMFJ, which reduced microbial complexity in industrial fermentation environments compared to natural ecosystems.

Fig. 5.

Dynamic Changes in alpha and beta diversity of microbial communities in douchi at different fermentation stages and species composition of douchi at different fermentation stages. (A) Ace and Chao indices; (B) Shannon and Simpson indices; (C) Principal component analysis; (D) Heatmap of distance matrix clustering based on bray-curtis statistics. (E) Species composition at the phylum level; (F) Species composition at the genus level; (G) Species composition at the species level.

β-diversity assessment integrating Principal Component Analysis (PCA) and sample distance analysis was employed to characterize structural dissimilarities in microbial communities between the YF and ZF. As shown in Fig. 5C, principal components PC1 (91.26%) and PC2 (5.61%) cumulatively explained 96.87% of the variance, indicating that PCA has a good ability to capture information. PCA plots (Fig. 5C) and clustering heatmaps (Fig. 5D) demonstrated a clear separation between the two sample types, along with significant variations in microbial composition across fermentation stages. The results demonstrate that Co-fermentation of YPKJ and YMFJ is the primary driving factor for microbial community difference.

3.4.2. Analysis of species composition in different types of fermented douchi

The microbial composition during the fermentation of YF and ZF reflects the succession patterns of microbial communities. At the phylum level (Fig. 5E), the predominant fungal phyla include Ascomycota, Mucoromycota, and Basidiomycota. Ascomycota consistently demonstrated a relative abundance exceeding 88%, establishing itself as the most predominant phylum across all fermentation stages. The relative abundance of Ascomycota in YF increased from 95.22% during the post-fermentation stage Y0 to 99.05% at Y20, followed by a decline to 88.76% at Y30, which is generally consistent with previous findings (Zhao et al., 2023). Notably, ZF exhibited an identical growth trend, with Ascomycota abundance at Z30 being 5.48% higher than that in YF. The relative abundance of Mucoromycota remained relatively low, averaging 4.12% in YF and less than 1% in ZF, which exhibited an initial decline followed by an increase in Mucoromycota abundance, with ZF consistently demonstrating lower levels throughout the fermentation process. In contrast to Ascomycota and Mucoromycota, Basidiomycota reached its maximum relative abundance at Y10 (1.11%) in YF and Z30 (3.33%) in ZF.

Fig. 5F shows that all 13 fungal genera were identified at the genus level (relative abundance >1%), and the fungal genera with higher relative abundance were Aspergillus, Millerozyma, Hyphopichia, and Pichia spp. At the species level, a total of 14 prominent fungal species (relative abundance >1%) were identified, primarily including Aspergillus flavus, Millerozyma farinosa, Hyphopichia burtonii, and Pichia kudriavzevii (Fig. 5G). The distribution patterns of these microbial species largely corresponded to their respective genus-level abundances. The Aspergillus spp. and Millerozyma spp. were identified as the dominant fungal taxa during different post-fermentation stages, corroborating previous research findings (Zhao et al., 2023). The Aspergillus spp. showed a decline in relative abundance from 87.77% (Y0) to 21.97% (Y30) during YF post-fermentation, though its average relative abundance remained ≥57.13%. In ZF, the Aspergillus spp. exhibited an abundance trend similar to that observed in YF, albeit with a significantly lower mean relative abundance (< 26.26%). The relative abundance of Millerozyma spp. showed a continuously increasing trend in YF, reaching its peak (59.82%) at the Y30 stage. However, Millerozyma spp. initially increased from 27.00% (Z0) to 85.19% (Z20), then declined rapidly to 65.13%, ultimately replacing Aspergillus spp. as the absolutely dominant species in the ZF. Furthermore, previous studies have confirmed that Millerozyma farinosa could promote the production of ethyl acetate, Isoamyl acetate, and 2-nonanol and have demonstrated the existence of esterification and dehydrogenation pathways (Jiang et al., 2021). Xu et al. (2023) demonstrated that inoculating Millerozyma farinosa as a fermentation starter in Daqu could promote microbial community succession and enhance the content of esters and aromatic compounds in strong-flavor Daqu. The Hyphopichia spp. exhibited significantly higher relative abundance in YF compared to ZF, peaking at 20.62% during the Y20 stage, but subsequently declined to below 1%. Meanwhile, Pichia spp. also demonstrated notable dynamics during douchi post-fermentation, in which abundance in YF increased from an initial 1.43% to 14.85%, reaching peak levels at Y10 before decreasing to <0.15% in later fermentation phases. Unlike YF, the Pichia spp. decreased from 5.50% in Z0 to 1.49% in Z10 during ZF fermentation, but its relative abundance then sharply increased to 17.58% by Z30. Xiao et al. (2023) inoculated Pichia spp. into chili peppers for fermentation and found that it, not only accelerated the fermentation process but also promoted the production of succinic acid and L-glutamic acid. Additionally, it generated various alcohols (such as phenethyl alcohol and 3-methyl-1-butanol) and phenols (including 4-ethyl-2-methoxyphenol and 2-methoxy-4-vinylphenol). Notably, Chen et al. (2025) demonstrated that Saccharomyces cerevisiae co-fermented with Pichia kudriavzevii of apricot wines significantly enriched ethyl acetate (fruity), ethyl caprylate (apple), and ethyl 2-methyl butanoate (sweet fruity).

In summary, inoculation with the two yeast starter cultures altered the microbial community structure and diversity, increasing the abundance of Millerozyma farinosa and Pichia kudriavzevii while reducing the relative abundance of Aspergillus spp.

3.5. Physicochemical index analysis

Table 1 shows the salt, total acidity, amino acid nitrogen, and reducing sugar contents of YF and ZF douchi during the fermentation process. The total acidity content of both YF and ZF increased with fermentation time, with YF rising from 1.41 to 2.58 g/100 g, while ZF increased from 1.33 to 2.59 g/100 g. This phenomenon could be attributed to the substantial accumulation of alkaline ammonia and other substances resulting from microbial enzymatic systems during fermentation and endogenous protein degradation (Shankar Ilango, 2021). With prolonged fermentation, the increase in amino acid nitrogen gradually stabilized, while ZF exhibited relatively higher levels, with the Z30 content being 2.3 times that of Z0. This phenomenon may result from protein hydrolysis mediated by various microorganisms, including yeast (Lan et al., 2023). Compared to YF, ZF exhibited more stable Salinity during post-fermentation, consistently measuring 8.65 g/100 g. By day 30 (Z30), its Salinity was 1.42 g/100 g lower than Y30, suggesting that the lower-salt environment could enhance yeast proliferation and metabolic activity. The reduced sugar content showed similar trends in YF and ZF, exhibiting an initial increase followed by a subsequent decrease. This pattern may be attributed to the further hydrolysis of D-galactose and glucose by β-galactosidase, which converts them into reducing sugars for microbial utilization and providing energy and nutrients for microbial growth and metabolism (Lan et al., 2023).

Table 1.

Changes in the basic physicochemical composition of YF and ZF fermented douchi at different fermentation stages.

Sample	Salinity	Total acidity	Amino acid nitrogen	Reducing sugar
	(g/100 g)
Y0	11.20 ± 0.14a	1.41 ± 0.08d	0.51 ± 0.03d	1.72 ± 0.08cd
Y10	9.60 ± 0.14d	2.08 ± 0.01c	0.87 ± 0.05c	2.13 ± 0.16cd
Y20	10.78 ± 0.16b	2.33 ± 0.04b	0.97 ± 0.07bc	3.84 ± 0.38a
Y30	10.20 ± 0.14c	2.58 ± 0.09a	1.10 ± 0.04a	1.63 ± 0.04d
Z0	8.85 ± 0.07e	1.33 ± 0.02d	0.50 ± 0.03d	2.11 ± 0.07cd
Z10	8.15 ± 0.07f	2.07 ± 0.04c	0.96 ± 0.04bc	3.36 ± 0.52ab
Z20	8.82 ± 0.01e	2.43 ± 0.01b	1.05 ± 0.02ab	2.73 ± 0.19bc
Z30	8.78 ± 0.04e	2.59 ± 0.04a	1.15 ± 0.07a	2.62 ± 0.28bc

Note: Values with different lowercase letters within a row/column differ significantly (p < 0.05). Y0, Y10, Y20, and Y30 are samples from the 5th day of koji-making and the 10th, 20th, and 30th days of post-fermentation of YF, respectively. Z0, Z10, Z20, and Z30 are samples from the salt-mixed inoculation before fermentation, and the 10th, 20th, and 30th days of post-fermentation of ZF, respectively.

3.6. Analysis of organic acid content

The organic acids in douchi may be produced through protein hydrolysis, microbial metabolic activities, and fermentation processes (He et al., 2019). In the study, four organic acids were identified in fermented douchi from YF and ZF, including succinic acid, malic acid, citric acid, and lactic acid. As shown in Fig. 6A, the succinic acid content in YF demonstrated initially decreased followed by an increase, reaching its peak at Y30 (2.28 ± 0.11 mg/g). Notably, the succinic acid content in ZF showed a progressive increase with prolonged fermentation time. Strikingly, the concentration at Z20 (2.30 ± 0.25 mg/g) exceeded that of YF at 30 d, ultimately reaching its peak level (3.02 ± 0.11 mg/g) by the 30d fermentation endpoint. The observed increase in succinic acid may be attributed to its role as an intermediate in the tricarboxylic acid (TCA) cycle, where microbial sugar metabolism drives its substantial secretion throughout fermentation (Jiang et al., 2025). Previous studies have demonstrated that succinic acid is a common organic acid in fermented soybean products and has been found to contribute to the umami flavor profile of douchi (Guo et al., 2024). Notably, lactic acid was the most abundant organic acid in fermented douchi. During YF post-fermentation, its content initially increased before decreasing, whereas ZF exhibited an opposite trend. This dynamic pattern may be associated with fluctuations in lactic acid-producing microorganisms under anaerobic fermentation conditions. Additionally, YF consistently exhibited higher lactic acid levels than ZF throughout all post-fermentation stages, potentially attributable to ZF's lower microbial diversity compared to YF. Fan, Chen, Du, and Fang (2024) conducted a correlation analysis between microorganisms and lactic acid during the fermentation process of strong-flavor baijiu, revealing a positive correlation between dominant yeast strains and lactic acid synthesis. YF and ZF exhibited fluctuating trends in malic acid content, but ZF showed higher levels at Z30 (0.52 ± 0.14 mg/g) compared to Y30, with an approximate difference of 0.10 mg/g. Previous studies have confirmed that yeast strains can produce malic acid through the sugar degradation of substrates (Teleky, Martău, Ranga, Chețan, & Vodnar, 2020). Compared to YF, ZF exhibited more stable trends in citric acid content, likely attributable to the more consistent fermentation performance of microbial communities in ZF. In summary, the two yeast strains used for Co-fermentation fermentation accelerated succinic acid production, which promoted succinic acid, malic acid, and citric acid in douchi, potentially contributing to flavor development.

Fig. 6.

Dynamic changes in organic acids and free amino acids during different fermentation stages of douchi and sensory evaluation analysis. (A) Organic acid content; (B) content of nucleotides (C) Free amino acid content; (D) Sensory evaluation score chart of YF and ZF.

3.7. Analysis of flavor nucleotide content

Flavor nucleotides are effective flavor enhancers composed of nucleotides, which are closely related to the taste of foods and exhibit a synergistic effect on flavor enhancement (Liu et al., 2024). In fermented products, flavor nucleotides promote the production of organic acids and free amino acids, thereby generating a pronounced umami taste (Nie et al., 2025). As shown in Fig. 6B, GMP and AMP were the primary umami contributors during douchi fermentation, maintaining relatively high concentrations throughout the process. GMP exhibited a linear increasing trend in YF, while it showed a fluctuating pattern of increase, followed by a decrease and subsequent rise and AMP also displayed fluctuations in ZF. This phenomenon is likely attributed to their interconversion mediated by microbial enzyme systems, which leads to dynamic fluctuations in the nucleotide profile. Furthermore, compared to YF, GMP and AMP were more prominent in ZF. GMP served as the primary contributor to the increased nucleotide content in fermented products, likely derived from adenosine triphosphate (ATP) degradation, and could enhance the umami taste of douchi (Wu, Xie, & Wang, 2021). GMP may synergistically interact with free amino acids such as glutamate (Glu) and phenylalanine (Phe) to promote umami enhancement (Qing et al., 2024). Throughout the douchi fermentation process, UMP and IMP maintained relatively low concentrations. However, their levels in Z30 were 1.93 mg/g and 1.64 mg/g higher than in Y30, respectively, which difference could be attributed to enhanced metabolic activity promoted by yeast strains during mixed fermentation.

3.8. Analysis of free amino acid content

Free amino acids (FAAs) serve as precursors for various volatile flavor compounds and represent a key indicator for evaluating the flavor quality of fermented soybean products (Guo et al., 2024). According to previous studies, FAAs have been identified as the primary nitrogen source for microbial growth and metabolism, playing a pivotal role in shaping the flavor characteristics of fermented foods (Song et al., 2024). In fermented douchi, all 17 free amino acids (FAAs) were identified (Fig. 6C), which could be categorized as umami-amino acids (Asp and Glu), sweet-amino acids (Ser, Gly, Thr, and Ala), sweet-bitter-amino acids (Pro and Lys), and tasteless-amino acids (Tyr and Cys) (Chen, Zhang, et al., 2024; Chen, Zhao, et al., 2024). As evidenced by Fig. 6C, ZF consistently exhibited higher total FAAs content than YF across all post-fermentation stages, with the Z10 timepoint surpassing Y30 by approximately 4.01 mg/g. In particular, the contents of Asp and Glu in ZF ranged from 1.45 to 5.43 mg/g and 2.88 to 8.82 mg/g, respectively, demonstrating a gradually increasing trend with prolonged fermentation time. These amino acids could synergize with Na⁺ to enhance the umami taste of douchi (Wang et al., 2025). However, YF exhibited relatively higher levels of bitter-tasting and umami amino acids, suggesting that bitter amino acids may coexist with Glu and potentially exert synergistic effects on umami perception (Yang et al., 2022). Furthermore, Sweet amino acids exhibited an initial rise followed by stabilization in ZF, whereas YF showed a gradual increasing trend, with overall levels remaining significantly higher in ZF. Interestingly, the contents of Ser and Ala were significantly higher ZF compared to YF. Overall, mixed-culture fermentation with two yeast strains could promote the production of free amino acids and may positively contribute to the flavor of fermented douchi.

3.9. Analysis of sensory evaluation results

Sensory evaluation is a key indicator for assessing the quality of fermented douchi, as it directly reflects their flavor characteristics and overall quality. Although ZF and YF received similar scores in appearance, ZF scored slightly higher, suggesting that ZF-fermented products may be more favored by evaluators in terms of visual attributes. As shown in the Fig. 6D, compared to YF, the inoculated fermentation (ZF) group achieved higher scores in all three sensory criteria: flavor, texture, and appearance. This could be due to the added bacterial strain effectively increasing the content of esters, phenols, and alcohols in the fermented douchi, thereby enhancing floral, mellow, and smoky notes. Meanwhile, the Co-fermentation process also facilitated the accumulation of free amino acids and umami nucleotides, which significantly improved the product's umami taste and overall flavor profile.

3.10. Dynamic changes of VOCs in douchi during co-fermentation

This study employed HS-SPME coupled with GC–MS to analyze the changes in volatile flavor compounds during the YF and ZF fermentation processes. The results revealed that all 156 VOCs were identified, including aldehydes (16), esters (47), ketones (12), alcohols (29), phenols (8), acids (16), hydrocarbons (15), furans (2), and other (4) compounds (Table S3). An Upset plot was employed to visualize both unique and shared VOCs among different douchi samples, with the results presented in Fig. 7A. A total of 57, 95, 93, and 95 VOCs were identified in YF at Y0, Y10, Y20, and Y30 fermentation stages, respectively, while 64, 98, 97, and 105 compounds were detected in ZF at corresponding Z0, Z10, Z20, and Z30 stages. The results demonstrated that VOCs in both types of douchi were differentially transformed by distinct microbial communities, leading to quantitative fluctuations. Notably, only 35 VOCs were shared across all eight sample groups. Furthermore, unique volatile compounds were identified in all eight sample groups, with YF samples containing 3 (Y0), 1 (Y10), 1 (Y20), and 4 (Y30) distinctive compounds, while ZF exhibited 4, 5, 8, and 7 unique compounds in Z0, Z10, Z20, and Z30. These results demonstrate significant temporal variations in ZF and YF, the compositional profile, and relative abundances of VOCs in douchi (p < 0.05, Fig. 7B). In the YF fermentation process, the total VOCs content increased significantly from 1777.20 μg/kg at Y0 to 14,356.32 μg/kg by Y30. In the ZF fermentation process, the total VOCs content increased from 2267.76 μg/kg at Z0 to 15,280.56 μg/kg by Z30, representing an approximately 6.7 times increase. Remarkably, the relative concentrations in ZF were consistently higher than those in YF at all corresponding fermentation stages. These results demonstrate that Co-fermentation with the two yeast strains significantly altered the diversity and relative abundance of VOCs in douchi, while effectively promoting aroma development.

Fig. 7.

Dynamic changes in volatile flavor compounds of douchi during different fermentation stages,cluster heatmap of volatile flavor compounds based on OAV ≥ 1, VIP score plot, and correlation network diagram between key volatiles and dominant microorganisms.(A) Venn diagram of the number of types; (B) Relative content; (C) OPLS-DA differential analysis plot; (D) Cross-validation plot by 200 permutation tests; (E) Cluster heatmap and VIP score plot; red was positive and blue was negative and higher VIP values correspond to greater flavor contributions; (F) Correlation network diagram, solid yellow lines represent positive correlations, and gray dashed lines signify negative correlations, with thicker lines indicating stronger relationships. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Esters typically associated with floral and sweet aromas, serve as characteristic flavor contributors in fermented foods due to their relatively low odor thresholds and significantly impact overall flavor profiles (Xu et al., 2022). In YF and ZF fermented douchi, ester compounds exhibited significant dynamic changes (p < 0.05) throughout all post-fermentation stages. On the 30th day of fermentation, YF identified all 21 esters, while ZF detected 11 more than YF, and their relative contents also showed fluctuating trends (Table S3 and Fig. 7B). The relative content of ester compounds in ZF showed an increasing trend with prolonged fermentation time (ranging from 112.495 to 3916.49 μg/kg), eventually stabilizing in the final stages. This phenomenon indicates that microorganisms such as yeast progressively convert alcohols or acids into ester compounds during fermentation, with the esterification reaction gradually reaching equilibrium over time. It is noteworthy that the relative content of ester compounds in Z10 was approximately 1.38 times higher than that in Y30. This phenomenon suggests that Co-fermentation with both yeast strains can accelerate ester production. The main ester compounds detected in fermented douchi included isoamyl acetate, ethyl benzoate, ethyl palmitate, ethyl 2-methylbutyrate, and ethyl acetate, most of which were ethyl esters a finding that is generally consistent with previous research findings. Among these compounds, the concentration of isoamyl acetate in ZF increased linearly from an initial 15.94 ± 0.85 μg/kg to 379.40 ± 1.69 μg/kg by day 30 of fermentation, showing a clear upward trend in relative content. Comparative analysis revealed a significantly higher relative content of isoamyl acetate in ZF compared to YF. Notably, ZF reached 359.09 ± 1.63 μg/kg by day 20 of fermentation, already surpassing YF's day-30 level of 165.42 ± 0.79 μg/kg. Additionally, a pronounced increase in ethyl 2-methylbutyrate concentration was observed in ZF douchi, escalating from 68.24 ± 1.63 μg/kg (Z0) to 361.18 ± 0.42 μg/kg (Z30), potentially resulting from the enhanced proliferation of Pichia kudriavzevii during late-stage fermentation that facilitated ester biosynthesis. The presence of abundant ethyl compounds could impart sweet, fruity, and oily flavors to fermented foods, while simultaneously mitigating undesirable odors from fatty acids and free amino acids (Guo et al., 2024).

During the fermentation process of fermented foods, the monosaccharides are first converted to pyruvate via glycolysis, which monosaccharides are first converted to pyruvate via glycolysis, which is transformed into alcohols and other volatile flavor compounds through a series of biochemical reactions. Additionally, unsaturated fatty acids can be oxidized to produce aldehydes, which are reduced to form alcohols. The alcohol content in ZF and YF showed fluctuating trends during the entire post-fermentation period, characterized by an initial rapid increase followed by a decrease and subsequent rebound (Table S3 and Fig. 7B). Phenethyl alcohol is a characteristic flavor compound in fermented soybean products, imparting a distinct floral aroma to the final products. This compound was consistently detected throughout the post-fermentation stage in ZF and YF, with particularly prominent levels in ZF, in which relative content increased from an initial Z0 (141.89 ± 0.71 μg/kg) to Z30 (885.08 ± 2.73 μg/kg), through a decreasing trend (Z20). This phenomenon may result from yeast-mediated esterification reactions, where organic acids react to form esters (such as ethyl palmitate), consequently reducing their precursor concentrations. Significantly, the relative content of phenylethyl alcohol in ZF at the Z10 stage (428.38 ± 1.29 μg/kg was approximately 176.19 μg/kg higher than that in YF at Y20. This process originates from the conversion of amino acids derived from inorganic nitrogen assimilation and protein degradation into glutamate and α-keto acids, which subsequently form higher alcohols (such as phenethyl alcohol and isobutanol) (Liu et al., 2023). 1-octen-3-ol maintained consistently high relative levels throughout the post-fermentation period in ZF and YF, with particularly prominent concentrations in YF. This unsaturated alcohol derived from linoleic acid oxidation exhibits characteristic mushroom and earthy odors, and due to its low odor threshold, it significantly influences the flavor profile of fermented foods (Wang, Jin, Zhang, Ahn, & Zhang, 2012).

Aldehydes, acids, ketones, phenols, and pyrazines, in addition to alcohols and esters, represent important VOCs in douchi. Microbial degradation of amino acids and retro-fermentation could generate aldehydes, which play a significant role in aroma enhancement (Kuang et al., 2022). A total of 14 aldehydes were identified in YF, which was 2 fewer than those detected in ZF (Table S3). Co-fermentation with both yeast strains significantly enhanced the types and relative content of aldehydes (Fig. 7B), promoting the accumulation of phenylacetaldehyde and 3-furaldehyde (Table S3). Notably, these compounds primarily contribute floral (Zhao et al., 2024) and sweet caramel-like aromas (Yao et al., 2024), respectively. The relative content of ketones had minimal overall impact on ZF, Co-fermentation reduced a key off-odor unsaturated alcohol (1-octen-3-ol) and reshaped the ketone profile while markedly increasing fruity/floral esters. The variation trend of organic acids during the post-fermentation phase was consistent between YF and ZF, exhibiting an initial increase followed by a decrease and subsequent rebound. This pattern likely results from microbial metabolism producing acids, which then undergo esterification reactions with alcohols to form esters. Consequently, by the 30th day of post-fermentation, the variety and relative content of esters became more prominent in ZF than YF. Strikingly, the relative content of isovaleric acid (fruity) (Zhang et al., 2024) in ZF remained consistently higher than in YF during the first 20 days of post-fermentation, and this demonstrates that Co-fermentation with the two yeast strains facilitates enhanced production of fruity aroma compounds. Phenols exerted a substantial impact on douchi characteristics, particularly through smoke components such as 2-methoxy-4-vinylphenol and guaiacol. Remarkably, the guaiacol content in Z30 reached 3213.22 ± 7.35 μg/kg, representing 7.21 times increase compared to Y30, which could contribute to ZF's more intense flavor profile.

The OAV is commonly used to evaluate the contribution of aroma compounds to the overall flavor profile, and it is widely accepted that compounds with an OAV ≥ 1 are considered to have a significant impact on aroma characteristics. 41 key VOCs with OAV ≥ 1 were identified from the YF and ZF, and the results are shown in Figs. 7E and Table S4. The 41 VOCs with OAV ≥ 1 include 11 esters, 1 alcohol, 6 ketones, 7 aldehydes, 1 acid, 2 pyrazines, 1 furan, 6 phenols, and 1 hydrocarbon. Additionally, compounds with OAV > 100 include ethyl isobutyrate, isoamyl acetate, ethyl 2-methylbutyrate, methyl 2-methyl butyrate, 3-methyl-1-butanol, decanal, phenylacetaldehyde, 2-nonenal, 1-octen-3-one, 2-ethyl-3,5-dimethylpyrazine, and guaiacol, indicating their role as key flavor-active substances in douchi. Compounds such as ethyl isobutyrate, isoamyl acetate, and ethyl 2-methylbutyrate are the primary contributors to sweet, banana-like, and fruity aromas in YF and ZF. Notably, the OAVs of ethyl isobutyrate, ethyl isovalerate, and ethyl 2-methylbutyrate in Z30 reached 17,640.50, 2529.33, and 6019.67, respectively, which were approximately 1.73, 2.29, and 4.49 times higher than those in Y30. Previous research has indicated that ethyl isobutyrate, ethyl isovalerate, and ethyl 2-methylbutyrate are key flavor compounds in douchi and contribute positively to its flavor profile (Jiang et al., 2026). The OAVs of ethyl isovalerate and ethyl 2-methylbutyrate were more prominent on the 20th day of ZF post-fermentation than those on the 30th day of YF post-fermentation. Chen, Zhang, et al. (2024) and Chen, Zhao, et al. (2024) conducted a study on the salt-free fermentation of wheat gluten, which discovered that Pichia spp. was capable of producing various ester compounds, including ethyl isovalerate, isoamyl acetate, and isobutyl acetate. Besides ester compounds, guaiacol is another key aroma component that contributes significantly to the characteristic aroma of ZF, imparting a distinctive smoky aroma to the fermented douchi (Zhao et al., 2023). 1-octen-3-one(mushroom), an unsaturated ketone compound with a low odor threshold, was predominantly detected in YF but disappeared in the Z30 fermentation stage of ZF, being replaced by esters and other compounds. This is likely due to the yeast-mediated reduction of 1-octen-3-one to 1-octen-3-ol through a reduction reaction, followed by its ester consistency with carboxylic acids to produce esters. Therefore, inoculation yeasts for Co-fermentation facilitate the formation of fruity, floral, and smoky aromas while enhancing the overall aroma profile of douchi and accelerating the generation of aromatic compounds.

Based on the above, in the overall flavor profile of douchi, ethyl isobutyrate, ethyl isovalerate, and ethyl 2-methylbutyrate (OAV > 2000) are key contributors, providing the core fruity and sweet aromas. Their high activity is associated with the ester-producing function of yeast (such as Pichia spp.). Guaiacol (OAV > 100) imparts a characteristic smoky aroma. Although 1-octen-3-one disappears due to conversion, its role as a precursor and its influence on flavor dynamics remain equally important.

To investigate the impact of mixed yeast fermentation on douchi flavor, differential analysis was performed on YF and ZF samples at four post-fermentation stages. The OPLS-DA results (Fig. 7C) demonstrate that the two principal components collectively explain 73.60% of the variance, confirming the model's good explanatory power and strong predictive capability. Additionally, distinct sample clustering was observed, particularly with YF and ZF occupying separate quadrants at 30 d of fermentation, indicating significant differences in flavor compounds both during the fermentation process and between the experimental groups. Furthermore, the reliability of the OPLS-DA model was validated through 200 permutation tests, yielding intercepts of R² = 0.1852 and Q² = −0.585 (Fig. 7D). These results confirm model robustness without overfitting. The differential marker compounds among the key VOCs of soy products were evaluated using VIP > 1 (Feng et al., 2024). Among the 41 key compounds with OAV ≥ 1, 15 compounds with VIP > 1 were screened out, which showed different change trends (Fig. 7E). Guaiacol and 4-ethylphenol covered the entire post-fermentation stage of ZF, and it reaches its peak on the 30th day of post-fermentation (VIP = 2.63), while 4-ethylphenol reaches its peak on the 20th day of fermentation (VIP = 2.28), but due to their relatively large VIP, they have a significant impact on the unique smoky and soy aromas of douchi. Their accumulation may be related to the demethylation reaction produced by yeast or the metabolism and transformation of phenylalanine. Phenethyl alcohol exhibits a more pronounced performance in ZF compared to YF, with a VIP value of 1.32, and plays a crucial role in contributing to the floral aroma of douchi. Isoamyl acetate and ethyl benzoate also showed significant differences throughout the post-fermentation process of douchi and were more pronounced throughout the entire fermentation stage of ZF. Ethyl ester is mainly produced by the esterification of ethanol with acid through the metabolism of microorganisms (such as Pichia spp.) (Zhao et al., 2022).

In conclusion, it could be inferred that the addition of yeast Co-fermentation promotes the production of VOCs such as guaiacol, 4-ethylphenol, phenethyl alcohol, isoamyl acetate, ethyl benzoate, and 3-methyl-1-butanol, which helps to enhance the smoky, floral, fruity and mellow aromas of douchi, and could also improve the unpleasant smell of unsaturated alcohols produced by unsaturated ketones.

3.11. The correlation between key flavor compounds and microbial communities

The creation of a correlation network diagram was aimed at visualizing the relationships between 10 fungal species and 41 key volatile compounds. The screening criteria were established with a correlation coefficient |R| ≥ 0.65 and statistical significance of p < 0.05 (Fig. 7F and Table S5). The results revealed that Millerozyma farinosa and Pichia kudriavzevii exhibited positive correlations with multiple esters, primarily including fruity aroma-associated compounds such as ethyl acetate and ethyl octanoate. By employing enhanced inoculation of Daqu fermentation, found that Millerozyma farinosa exhibited positive correlations with ethyl acetate and ethyl octanoate. Millerozyma farinosa exhibited a positive correlation with phenethyl alcohol, which contributes to the rose-like aroma in douchi. Previous studies on fermented foods have confirmed that Millerozyma farinosa is the dominant microbial species responsible for encoding aldehyde dehydrogenase genes, which directly catalyze the conversion of phenylacetaldehyde to phenylethyl alcohol through the corresponding aryl alcohol dehydrogenase-coding gene (Yuan et al., 2023). The compounds acetoin and guaiacol also exhibit a positive correlation with Pichia kudriavzevii, which could enhance the creamy and smoky aromas characteristics in douchi. Notably, Millerozyma farinosa and Pichia kudriavzevii exhibit negative correlations with the unsaturated alcohol 1-octen-3-ol (mushroom) and the long-chain alcohol 2-octanol (Herba), respectively. Therefore, it could be speculated that Co-fermentation fermentation with both yeast strains promotes the formation of fruity, floral, creamy, and smoky aromas, while ameliorating the unpleasant odors caused by unsaturated alcohols.

4. Conclusion

This study successfully isolated and characterized two indigenous aroma-producing yeasts, Pichia kudriavzevii (YPKJ) and Millerozyma farinosa (YMFJ), from Yangjiang douchi. The former exhibited strong lipid synthesis capabilities, while the latter primarily produced alcohols and ketones. Through single-factor optimization followed by co-culture, the yeasts were found to alter the succession of the core fungal community: YPKJ accounted for 17.58% in the Z30 medium, while the relative abundance of YMFJ significantly increased to 85.19% at Z20, accompanied by suppression of Aspergillus spp. This targeted microbial shift drove notable metabolic reprogramming, leading to substantial increases in key taste components. Specifically, succinic acid peaked at 3.02 ± 0.11 mg/g at the Z30 stage, along with marked increases in key umami components such as free amino acids and nucleotides. Analysis by HS-SPME-GC–MS and OAV revealed substantial rises in fruity-floral esters and smoky phenolic compounds, while off-flavors such as 1-octen-3-ol decreased. As a result, the Co-fermentation process imparts a distinctive ‘fruity-floral-smoky’ aroma profile to Yangjiang fermented douchi while enhancing umami perception. These findings demonstrate that Co-fermentation with aroma-producing yeasts effectively regulates microbial communities and flavor development in high-salt fermented douchi, providing a theoretical foundation and a practical strategy for industrial standardization and flavor stability optimization. Furthermore, based on the preliminary regulatory effects and research value of Co-fermentation, future studies will incorporate strain-specific qPCR or genotyping techniques to precisely quantify the dynamic changes and interactions between the two yeast strains. This will provide more direct molecular evidence for establishing the causal relationship among “microbial community structure - metabolic function - flavor formation”.

CRediT authorship contribution statement

Donghui Luo: Writing – review & editing, Writing – original draft, Methodology, Data curation. Xiaoli Jiang: Writing – review & editing, Writing – original draft, Methodology, Data curation. Mengting Zhang: Validation, Conceptualization. Zilong Zhao: Formal analysis. Jianxu Chen: Visualization, Resources. Jian Liang: Visualization, Resources. Jingjing Guan: Validation, Supervision, Software, Resources. Huiyan Zhao: Writing – review & editing, Supervision, Funding acquisition, Conceptualization.

Funding

This work was supported by the Characteristics of Resource Development and Industrialization in Shangyang Town, Yangxi County (SDZX2022038), the·Scientific Research·Start-up Funds of Guangdong·Ocean University (360302062402), and Yangjiang City Weidu Condiment Research Laboratory (2024024).

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 data 1

GC-MS Flavor Profiling and Bacterial Strains.

Data availability

Data will be made available on request.