Dynamic changes in microbial communities and flavor during different fermentation stages of proso millet Baijiu, a new product from Shanxi light-flavored Baijiu

Introduction Proso millet, a high-quality fermentation material used for Chinese yellow wine production, can produce special flavored substances; however, its role in improving the flavor and altering microbial communities of light-flavored Baijiu during fermentation remain unknown. Thus, we aimed to investigate the effect of proso millet on improving the flavor of light-flavored Baijiu and altering microbial communities during different fermentation stages. Methods The dynamic changes in the microbial communities and flavor of proso millet (50%) + sorghum (50%) mixed fermentation samples were analyzed through intermittent sampling on days 7, 14, 21, and 28 of the fermentation process. Microbial high-throughput sequencing and the analysis of flavor characteristics were conducted through 16S DNA/ ITS amplicon sequencing and gas chromatography (multi-capillary column)-ion mobility spectrometry, respectively. Results Proso millet significantly changed the core flavor compound composition of traditional light-flavored Baijiu from ethyl acetate, ethyl hexanoate, ethyl hexanoate dimer, ethyl butanoate, ethyl lactate, and butyl acetate to oct-2-ene, 2-butanol, propyl propanoate, 2-pentenal, and 4-methylpentanal. The amplicon sequencing analysis revealed that the alpha diversity parameters of bacterial and fungal communities, including the Chao1, Pielou_e, Shannon, and Simpson indices, for proso millet–sorghum mixed fermentation samples were significantly higher than those for sorghum fermentation samples (p < 0.05). Of the 40 most significant microbial genera in two treatments, proso millet significantly increased the abundance of 12 bacterial and 18 fungal genera. Among the 40 most significant bacterial and fungal species, 23 bacterial species belonged to the Lactobacillus genus, whereas the 30 primary fungal species belonged to 28 different genera. The analysis of the relationship between microbial changes and the main flavor compounds of light-flavored Baijiu showed that bacteria from the Weissella, Acinetobacter, Bacteroides, Psychrobacter, Pseudarthrobacter, Lactococcus, Chloroplast, Saccharopolyspora, Psychrobacter, Saccharopolyspora, Pseudonocardiaceae, Bacteroides genera and fungi from the Thermoascus, Aspergillus, Pichia, Rhizomucor, Papiliotrema, Hyphopichia, and Mucor genera significantly inhibited the synthesis of ethyl hexanoate, ethyl butanoate, ethyl lactate ethyl lactate, and butyl acetate but increased the synthesis of ethyl acetate (p < 0.05). Moreover, these microbes exhibited a significantly greater abundance in proso millet–sorghum mixed fermentation samples than in sorghum samples. The synthesis of special flavored compounds in proso millet Baijiu was significantly positively correlated with the presence of fungi from the Rhizopus, Papiliotrema, Wickerhamomyces, Aspergillus, and Thermoascus genera but negative correlated with the presence of bacteria from the Weissella, Acinetobacter, Psychrobacter, Pseudarthrobacter, Bacteroides, and Saccharopolyspora genera. Regarding ethanol content, the low alcohol content of Fenjiu may be due to the significantly high abundance of fungi from the Psathyrella genus and bacteria from the Staphylococcus, Kroppenstedtia, Brevibacterium, and Acetobacter genera during fermentation. In summary, proso millet significantly altered the flavor of light-flavored Baijiu by inducing the formation of a special microbial community; however, it did not increase alcohol concentration. Discussion This study lays the foundation for future research on Baijiu fermentation. Additionally, the study findings may help improve the production efficiency and elevate the quality and flavor of the final product.


Introduction
In China, Baijiu has evolved from being a mere beverage to a symbol of culture.The Baijiu culture in China has a rich historical background, and companies that produce distilled Baijiu operate on a large scale with huge profits.Fenjiu, traditional Chinese Baijiu primarily made from non-glutinous sorghum, is highly popular light aroma-style beverage that has been consumed for over 1,500 years (Ma et al., 2013).However, companies that produce light-flavored Baijiu face several challenges, including low yield, limited high-end product options, declining quality, and a stagnant market due to a narrow flavor profile, which have hindered their development (Yifan et al., 2021).Currently, researchers are actively exploring methods to enhance their understanding of light-flavored Baijiu and improve its quality, with research focusing on aspects such as the microbial community structure in light-flavored Daqu, flavor chemistry of light-flavored Baijiu, influence of sorghum varieties on the flavor of light-flavored Baijiu, and the relationship between light-flavored Baijiu and microorganisms (Liu and Sun, 2018;Huang et al., 2020;Zhang et al., 2022;Xiang et al., 2023;Zhang et al., 2023).In addition, studies have examined the physicochemical properties of Chinese light-flavored Baijiu during storage and explored ways to improve sorghum varieties for Baijiu production (Ma et al., 2013;Kang et al., 2020).Furthermore, to improve the quality of light-flavored Baijiu, some other traditional brewing materials such as rice, barley, and glutinous sorghum, have been currently added to the raw materials used for its production.However, it remains unclear whether proso millet, which is the primary raw material for Chinese yellow wine production, can improve the flavor of light-flavored Baijiu.In addition, the microbial communities and flavor profiles of sorghumbased light-flavored Baijiu during fermentation have not yet been determined.
Proso millet (Panicum miliaceum L.) is an annual herbaceous cereal crop, commonly known as "huangmi" in North and Northwest China; it has a high nutritional value and is rich in starch, dietary fibers, proteins, phenolic compounds, vitamins, amino acids, and trace elements (Kalam Azad et al., 2019;Wiedemair et al., 2020;Yuan et al., 2022).It is extensively cultivated in arid or semi-arid regions across North China, Asia, Australia, North America, Europe, and Africa (Sanderson et al., 2017;Yang et al., 2018;Kučera et al., 2019;Gong et al., 2021;Narciso and Nyström, 2023).Moreover, proso millet has demonstrated antioxidant, anti-proliferative, antimicrobial, and anti-carcinogenic properties.It has also been associated with alleviating celiac disease, regulating blood glucose and cholesterol levels, and preventing diabetes and cardiovascular disorders (Kim et al., 2011;Sanderson et al., 2017;Shen et al., 2018;Yang et al., 2018;Kalam Azad et al., 2019;Agarwal and Chauhan, 2019).Consequently, proso millet is a valuable resource for the development of healthy foods, and has recently been used to produce popular products, such as flour, bread, couscous, biscuits, porridge, extruded snacks, and fermented beverages (Sanderson et al., 2017;McSweeney et al., 2017a;Wang J. et al., 2020).To promote the development of new products, increase awareness regarding the functional properties of proso millet, and determine the potential applications of proso millet in pasta processing or other cooking processes, significant research attention has been devoted to topics such as drought resistance, cultivation methods, yield improvement, breeding, functional properties, physicochemical characteristics, nutritional composition, protein properties, starch physicochemical properties, and digestibility of the crop, as well as its biological structure and microstructure analyses (Gulati et al., 2017(Gulati et al., , 2018;;Hasseldine et al., 2017;McSweeney et al., 2017b;Balkrishna and Visvanathan, 2019;Johnson et al., 2019;Yang et al., 2019;Liu et al., 2020;Wiedemair et al., 2020;Gong et al., 2021;Chang et al., 2023;Kumar et al., 2023).However, limited research has been carried out on its fermentation characteristics and its use in the development of fermented products.Despite the development of healthy proso millet wines (yellow wine or fermented alcoholic drinks), beer, and vinegar, with recent studies shedding light on their flavor composition and fermentation processes (Liu et al., 2018;Santra et al., 2019;Zheng et al., 2020;Bian et al., 2022), there is limited research on new fermentation products, particularly hard liquor made from proso millet.Moreover, studies on the relevant fermentation 10. 3389/fmicb.2024.1333466Frontiers in Microbiology 03 frontiersin.orgprocesses, flavor profiles, and mechanisms underlying quality formation for proso millet liquor are limited.Therefore, in recent years, our research team has optimized the fermentation formula and technology based on Fenjiu Daqu, leading to the successful development of a new product (ZL201410293583.1).In this study, we aimed to investigate the effects of proso millet on improving the flavor of light-flavored Baijiu by studying the microbial composition and flavor differences between the new product and sorghum-based light-flavored Baijiu during fermentation.In addition, we aimed to reveal the mechanistic correlations between microbial community composition and flavor formation.Our findings may help expand the application range of proso millet and provide a reference basis for improving the flavor of Fenjiu or other light-flavored Baijiu beverages.

Experimental design and sampling
Sorghum (Jinza No. 22) and proso millet (Jinshu No. 9) were procured from Shanxi Jinliang Agricultural Development Company, Xiyang County, Jinzhong City, Shanxi Province, China.Two fermentation tests, sorghum and mixed material (composition of 50% sorghum and 50% proso millet) fermentation, were conducted under commercial conditions at the Xinghua Village Fen Distillery in Shanxi Province.Samples of fermented grains (FG) were collected at four different time points during alcoholic fermentation in the pit: days 7, 14, 21, and 28.To ensure comprehensive data and representativeness, each sample at every time point comprised nine subsamples collected from various spatial positions within the underground vats, with each subsample weighing 100 g (Supplementary Figure S1).Three parallel samples (from three random cellars) were collected for each sample type.Subsequently, all 24 samples were transported to the laboratory on ice and stored at −80°C until further analysis.The mixed proso millet fermentation samples collected on days 7, 14, 21, and 28 of fermentation were designated as MA, MB, MC, and MD, respectively.The sorghum fermentation samples collected on days 7, 14, 21, and 28 of fermentation were designated as GA, GB, GC, and GD, respectively.

DNA extraction and high-throughput sequencing analysis
Total genomic DNA was extracted from 200 mg of FG samples using Fast DNA SPIN extraction kits (MP Biomedicals, Santa Ana, CA, USA), following the manufacturer's protocol.The quantity and quality of the extracted DNA were assessed using an ND-1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and agarose gel electrophoresis, respectively.
PCR amplicons were purified using Agencourt AMPure Beads (Beckman Coulter, Indianapolis, IN, USA) and quantified using the PicoGreen dsDNA Assay Kit (Invitrogen, Carlsbad, CA, USA).Following individual quantification, amplicons were pooled in equal amounts, and pair-end 2 × 300 bp sequencing was conducted on the Illumina MiSeq platform using MiSeq Reagent Kit v3, carried out by Shanghai Personal Biotechnology Co., Ltd.(Shanghai, China).

Bioinformatics and statistical analysis
The Quantitative Insights Into Microbial Ecology (QIIME, v1.8.0) pipeline was employed to process sequencing data, following established procedures (Caporaso et al., 2010).In summary, raw sequencing reads that precisely matched the barcodes were assigned to their respective samples and identified as valid sequences.Paired-end DNA fragments were initially assembled using FLASH (Zhang et al., 2014).After chimera detection, the remaining highquality sequences were clustered into operational taxonomic units (OTUs) with 97% sequence identity using UCLUST (Jiang et al., 2012).Taxonomic assignments for representative sequences from each OTU were determined using the Silva database (16S rDNA) and Unite database (ITS) within the QIIME platform (Caporaso et al., 2010).Data analysis was primarily conducted using the QIIME and R packages (v3.2.0).Alpha diversity indices at the OTU level, including the Chao1 richness estimator, the abundance-based coverage estimator metric, the Shannon diversity index, and the Simpson's index, were calculated using the OTU table in QIIME.Differences in microbiota structure among groups were assessed using permutational multivariate analysis of variance and analysis of similarities using the "vegan" function in R (Takeshita et al., 2012;Kelly et al., 2015).Taxonomic compositions and abundance were visualized using MEGAN and GraPhlAn (Li et al., 2019).Taxa abundances at the phylum, class, order, family, genus, and species levels were statistically compared among samples or groups using Metastats (Elokil et al., 2020) and visualized using violin plots.Raw sequence data have been deposited in the NCBI Sequence Read Archive under the accession numbers, PRJNA1015004, PRJNA1014983, PRJNA1013705, and PRJNA1014132.

Flavor analysis
The FG samples used for flavor analysis were the same as those mentioned in Section 2.1.Volatile compounds were detected using FlavourSpec ® (G.A.S., Germany) with an FS-SE-54-CB-1 column (15 m length × 0.53 mm inner diameter) via GC-IMS.Briefly, 2 g of 10.3389/fmicb.2024.1333466Frontiers in Microbiology 04 frontiersin.orgeach fermented sample was placed in a 20-mL headspace bottle, incubated at 60°C, and centrifuged at 500 rpm for 15 min.Subsequently, 200 μL of liquid was extracted from the headspace bottle using an 85°C syringe and automatically injected into the injector of the GC-IMS equipment.Nitrogen gas with 99.999% purity was used as the carrier gas, and volatile organic compounds (VOCs) in the FG samples were separated using a 60°C quartz capillary column.The GC flow conditions were as follows: 2 mL/min for 2 min, 100 mL/min for 20 min, 150 mL/min for 30 min, and then the flow was halted.VOC separation was carried out at a column temperature of 60°C, with detection at 45°C in the IMS ionization chamber at a drift gas flow rate of 150 mL/min.All experiments were conducted in triplicate.

Statistical analyses
Statistical analyses were performed using one-way analysis of variance in the SPSS version 17.0 software (SPSS, Inc., Chicago, IL, USA).Differences between groups were evaluated using Duncan's multiple comparison tests, and p < 0.05 was considered statistically significant.

Overall analysis of Illumina MiSeq data
In total, 324,421 and 292,751 high-quality sequences were obtained from the bacterial 16S rRNA V3-V4 sequences, whereas 514,202 and 470,916 high-quality sequences were obtained from the fungal ITS sequences for sorghum and proso millet FG samples, respectively.These sequences were clustered into 3,133 and 371 (sorghum FG samples) and 7,255 and 574 (proso millet FG samples) amplicon sequence variants (ASVs), respectively, at a 97% similarity level.The coverage estimator for bacterial and fungal ASVs exceeded 99% for all samples, indicating that the obtained sequence reads adequately represented the microbial diversity.
Microbial community richness, diversity, and evenness in sorghum and proso millet FG samples were assessed using the Chao1, Shannon, Simpson, and Pielou's evenness indices (Table 1).Microbial diversity in both sorghum and proso millet FG samples decreased as fermentation time increased.Despite showing a lower number of high-quality sequences, proso millet FG samples exhibited a higher microbial diversity than sorghum FG samples as the bacterial Chao1, Shannon, Pielou's evenness, and Simpson indices for proso millet FG samples at different stages of fermentation were approximately twice as those for sorghum FG samples; in addition, the fungal Chao1 index was considerably higher for proso millet FG samples than for sorghum FG samples (Table 1).

Flavor compound composition of proso millet Baijiu fermentation samples
As shown in Figure 3, there were significant differences among the four samples (MA-MD).The dominant compositions of flavor compounds and their abundances in the four samples were also different.Moreover, compared with MA samples, MD samples exhibited evident differences in the abundances of compounds (Figure 4).
In addition, as shown in Figure 7, the abundance of compounds in areas 1 and 3 was higher than that in other samples, and these compounds could be used as characteristic flavor compounds in MA and MD samples, respectively.In area 2, the levels of compounds such as ethyl propionate, ethyl lactate, acetic acid, and ethyl butyrate gradually increased and attained a maximum abundance during fermentation.In area 4, compound abundance remained relatively constant during fermentation.Flavor compounds in MA and MD samples were significantly different, whereas those in MB and MC samples were relatively similar.In area 5, the abundance of compounds such as butyl acetate, 3-hydroxy-2-butanone, and butyraldehyde was significantly higher in GA samples than in GB-GD samples.In area 6, compound abundance initially increased and then decreased.However, in area 7, the abundance of compounds such as ethyl 3-methylbutyrate, ethyl caproate, ethyl butyrate, ethyl propionate, ethyl lactate, propyl acetate, isoamyl acetate, and isobutyl acetate, gradually increased during fermentation.In area 8, compound abundance remained relatively constant throughout the fermentation process.Difference diagram showing the gas phase ion mobility spectrum of the MA, MB, MC, and MD samples.The MA sample was selected as the reference.The spectra of the other samples were obtained from a previous study.If two volatile organic compounds are the same, the deducted background is white, red indicates that the concentration of the substance is higher than that of the reference, and blue indicates that the concentration of the substance is lower than that of the reference.

Relationship between microbial composition changes and flavor compound formation
The RDA analysis result showed that bacteria from the Weissella, Acinetobacter, Bacteroides, Psychrobacter, Pseudarthrobacter, Lactococcus, Chloroplast, Saccharopolyspora, Psychrobacter, Saccharopolyspora, Pseudonocardiaceae, and Bacteroides genera and fungi from the Thermoascus, Aspergillus, Pichia, Rhizomucor, Papiliotrema, Hyphopichia, and Mucor genera significantly inhibited the synthesis of the flavor compounds, such as ethyl hexanoate, ethyl butanoate, ethyl lactate ethyl lactate, and butyl acetate, but improved the abundance of ethyl acetate (p < 0.05) (Figures 8, 9).Moreover, these microbes exhibited a greater abundance in proso milletsorghum mixed fermentation samples than in sorghum samples.The synthesis of special flavor compounds in proso millet Baijiu was significantly promoted in the presence of fungi from the Rhizopus, Papiliotrema, Wickerhamomyces, Aspergillus, and Thermoascus genera but inhibited by bacteria from the Weissella, Acinetobacter, Psychrobacter, Pseudarthrobacter, Bacteroides, and Saccharopolyspora genera (Figures 8, 9).
Among the 40 fungal and bacterial species with the highest abundance, all bacterial and 30 fungal species were found to be significantly associated with changes in the abundance of flavor compounds (Supplementary Table S2).More specifically, among these 30 fungal species, P. fermentans, C. tropicalis, D. geotrichum, S. inaequale, P. membranifaciens, C. curvatus, C. ethanolica, and R. arrhizus played significant roles in altering the abundance of flavor compounds, and were significantly correlated with changes in the abundance of 18, 18, 18, 15, 13, 11, 10, and 9 types of flavor compounds, respectively.In addition, eight main fungal species exerted a significant effect on changes in the abundance of 41 compounds, with the exception of the ethyl 3-methyl butanoate monomer, the ethyl 3-methyl butanoate dimer, 2-methylfuran, and thiophene.The ethyl 3-methyl butanoate monomer and the ethyl 3-methyl butanoate dimer were found to be significantly positively correlated with the presence of F. solani and P. candolleana (p < 0.05).In addition, the abundance of 2-methylfuran was significantly positively correlated with F. solani abundance (correlation coefficient R = 0.475), but negatively correlated with R. arrhizus abundance (correlation coefficient R = −0.443)(p < 0.05).Thiophene abundance was significantly positively correlated with P. fermentans, T. pullulans, and C. ethanolica abundance (p < 0.01), with correlation coefficients of 0.701, 0.600, and 0.741, respectively; however, thiopene abundance was negatively correlated with G. baccata abundance (p < 0.05) (correlation coefficient R = −0.431).

Discussion
Proso millet, a small miscellaneous grain rich in β-carotene and multiple vitamins, is often utilized in the production of rice wine and other viscous foods due to its high viscosity.However, its application in Chinese Baijiu fermentation faces significant challenges.In a previous study, to effectively utilize millet in Chinese Baijiu fermentation, we optimized the fermentation process and successfully developed a new Chinese Baijiu product.Moreover, sensory evaluation indicated significant differences in taste and flavor between  Gallery plot of volatile flavor compounds based on the gas phase ion mobility spectrum at different stages of proso millet and sorghum fermentation.
The numbers represent unidentified compounds in the mobility library.

FIGURE 8
Relationship between 45 flavor compounds and microbial communities at the genus level. 10.3389/fmicb.2024.1333466 Frontiers in Microbiology 14 frontiersin.orgthe new Chinese Baijiu product and traditional variants.Therefore, this study aimed to describe differences between traditional and new Chinese Baijiu products through the analysis of microbial communities and flavor at various fermentation stages.

Relationship between core microbes and flavor compound formation
The RDA analysis showed that proso millet induced a change in the core flavor compounds of sorghum Baijiu by increasing the abundance of flavor-related microbes, such as bacteria from the Weissella, Acinetobacter, Bacteroides, Psychrobacter, Pseudarthrobacter, Lactococcus, Chloroplast, Saccharopolyspora, Psychrobacter, Saccharopolyspora, Pseudonocardiaceae, and Bacteroides genera and fungi from the Thermoascus, Aspergillus, Pichia, Rhizomucor, Papiliotrema, Hyphopichia, and Mucor genera, the abundance of which was significantly greater in proso millet-sorghum mixed fermentation samples than in sorghum samples.In addition, among the core fungal species, F. solani was significantly positively correlated with the formation of the main flavor compound, A20.Furthermore, P. candolleana was significantly positively correlated with the formation of the main flavor compound, B3. S. inaequale was significantly negatively correlated with the formation of the main flavor compound, A9, but was significantly positively correlated with the formation of the main flavor compounds, B3, B4, B6, F1, and F2.Furthermore, T. aurantiacus and M. tassiana were significantly negatively correlated with the formation of the main flavor compound, E2.In addition, A. ruber was significantly positively correlated with the formation of the main flavor compounds, A11 and B9.R. arrhizus was significantly negatively correlated with the formation of the main flavor compound, C1, but was significantly positively correlated with the formation of the main flavor compound, F1.H. nigrescens was significantly negatively correlated with the formation of the main flavor compounds, E2 and G2.Furthermore, D. geotrichum was significantly negatively correlated with the formation of the main flavor compounds, A9, A15, A18, and A20, but was significantly positively correlated with the formation of the main flavor compounds, A19, B4, B6, and F2. C. tropicalis was significantly negatively correlated with the formation of the main flavor compounds, A9, A14, A18, A20, C1, G1, and G2, but was significantly positively correlated with the formation of the main flavor compounds, B4, B6, F1, and F2.T. dohaense was significantly negatively correlated with the formation of the main flavor compounds, A15 and E2.M. circinelloides was significantly negatively correlated with the formation of the main flavor compounds, A14, A15, and G2. S. podzolica was significantly negatively correlated with the formation of the main flavor compound, G2, but was significantly positively correlated with the formation of the main flavor compound, F1.G. baccata and A. fumigatus were significantly positively correlated with the formation of the main flavor compound, G1.L. ramosa was significantly negatively correlated with the formation of the main flavor compound, G2.O. truncatum was significantly positively correlated with the formation of the main flavor compounds, B5 and B9.F. magnum was significantly negatively correlated with the formation of the main flavor compounds, A19 and B3.Furthermore, B. atrogriseum was significantly negatively correlated with the formation of the main flavor compound, B3, but was significantly positively correlated with the formation of the main flavor compound, A14.P. fermentans was significantly negatively correlated with the formation of the main flavor compounds, A14, B9, and G1, but was significantly positively correlated with the formation of the main flavor compounds, A19 and G2. C. curvatus was significantly negatively correlated with the formation of the main flavor compound, A11, but was significantly positively correlated with the formation of the main flavor compound, E2.T. pullulans was significantly negatively correlated with the formation of the main flavor compound, A14.Among the core bacterial species, L. acetotolerans was significantly negatively correlated with the formation of the main flavor compounds, A19, B4, B6, E2, F1, and F2, but was significantly positively correlated with the formation of the main flavor compounds, A9, A15, A18, A20, and C1.L. homohiochii was significantly negatively correlated with the formation of the main flavor compounds, A19, B3, B4, B6, E2, F1, and F2, but was significantly positively correlated with the formation of the main flavor compounds, A9, A14, A15, A18, A20, C1, and G1.P. pentosaceus was significantly positively correlated with the formation of the main flavor compounds, A14, A15, B9, and G1.P. parvulus was significantly positively correlated with the formation of the main flavor compounds, A9, A19, A20, B4, B6, E2, and F1.P. damnosus was significantly negatively correlated with the formation of the main flavor compounds, A9, A19, and C1, but was significantly positively correlated with the formation of the main flavor compounds, A19, B4, B6, E2, and F2.L. coryniformis was significantly negatively correlated with the formation of the main flavor compounds, A9, A14, A15, A18, A20, B5, B9, and C1, but was significantly positively correlated with the formation of the main flavor compounds, A19, B4, B6, E2, F1, and F2.L. oligofermentans was significantly negatively correlated with the formation of the main flavor compounds, A14, A15, A18, A20, B5, B9, C1, and G1, but was significantly positively correlated with the formation of the main flavor compounds, A9, A19, B4, B6, E2, F1, and F2.L. buchneri and B. vulgatus were significantly positively correlated with the formation of the main flavor compounds, A11, B5, and B9.Furthermore, L. malefermentans and L. hokkaidonensis were significantly positively correlated with the formation of the main flavor compounds, A11, A14, B5, B9, and G1.L. carnosum was significantly positively correlated with the formation of the main flavor compound, A11, but was negatively correlated with the formation of the main flavor compound, C1.P. inopinatus was significantly positively correlated with the formation of the main flavor compounds, A11, B9, and G1.E. durans was significantly positively correlated with the formation of the main flavor compounds, A19, B4, B6, F1, and F2, but was negatively correlated with the formation of the main flavor compounds, A9, A18, A20, and C1.L. vaccinostercus and L. mesenteroides were significantly positively correlated with the formation of the main flavor compounds, A11, B3, B4, B6, F1, and F2, but were negatively correlated with the formation of the main flavor compounds, A9, A18, A20, and C1.L. mali was significantly positively correlated with the formation of the main flavor compounds, B3, B4, B6, F1, and F2, but was negatively correlated with the formation of the main flavor compounds, A9, A18, A20, C1, and G2.L. suebicus was significantly positively correlated with the formation of the main flavor compounds, B3, B6, F1, and F2, but was negatively correlated with the formation of the main flavor compounds, A9, A18, A20, C1, and G2.L. sakei was significantly positively correlated with the formation of the main flavor compounds, B4, B6, F1, and F2, but was negatively correlated with the formation of the main flavor compounds, A9, A15, A18, A20, and C1.W. viridescens, W. paramesenteroides, L. fermentum, L. paralimentarius, and L. fallax were significantly positively correlated with the formation of the main flavor compounds, B4, B6, F1 and F2, but were negatively correlated with the formation of the main flavor compounds, A9, A15, A18, A20, C1, and G2.

Effect of proso millet on ethanol production and the abundance of key microbial communities for ethanol synthesis
Ethanol concentration is a direct reflection of the Baijiu yield and is significantly affected by sorghum variety and the microbial communities present during fermentation.A comparison of sorghum fermentation and mixed fermentation showed no significant difference in ethanol concentration at the final fermentation stage 10.3389/fmicb.2024.1333466Frontiers in Microbiology 17 frontiersin.org(28 days); however, ethanol concentration at the fermentation stages on days 14 and 21 were significantly lower for sorghum fermentation than for mixed fermentation (Table 2).The RDA analysis results showed that the presence of fungal species of the genera Olipidium was significantly positively correlated with ethanol production; however, the presence of fungi from the Psathyrella and Schizothecium genera was significantly negatively correlated with ethanol production (Supplementary Table S2).In total, 12 bacterial genera were significantly negatively correlated with ethanol production, with the negative correlation coefficient for the genus, Brachybacterium, being higher than that for other genera (−0.582; p < 0.01).In addition, the relative abundance of the bacteria of the genus, Olipidium, in mixed fermentation samples was lower than that in sorghum samples; however, the relative abundance of microbes that inhibit ethanol synthesis was higher in mixed fermentation samples than in sorghum samples (Figures 2, 3).Therefore, we speculated that proso millet could improve the flavor of light-flavored Baijiu but could not improve its yield.Furthermore, L. buchneri, P. inopinatus, L. malefermentans, L. hokkaidonensis, P. pentosaceus, B. vulgatus, A. ruber, and O. truncatum played important roles in ethanol production in proso millet Baijiu.Conversely, during the production of light-flavored Xiaoqu Baijiu, ethanol production was significantly positively correlated with the presence of the microbe, Saccharomyces cerevisiae, and negatively correlated with the presence of A. pasteurianus (Xie et al., 2021).L. ramosa, Saccharomycopsis fibuligera, B. licheniformis, S. cerevisiae, and Pichia kudriavzevii play significant roles in starch degradation and ethanol production (Huang et al., 2020).Furthermore, S. cerevisiae, Z. bailii, and S. pombe were found to play important roles in ethanol production in Daqu Chinese Baijiu (Wu et al., 2021); in addition, S. cerevisiae was the main ethanol-producing microbe identified during zaopei fermentation (Zou et al., 2018).While the effects of proso millet on dynamic changes in microbial communities and flavor during different stages of light-flavored Baijiu fermentation using the light-flavored Baijiu technology of Xinghua Village, Shanxi, have been studied and some key microorganisms closely related to flavor changes identified, several important aspects still require further research.These include determining the optimal quantity of proso millet to be added, selecting superior proso millet varieties, investigating the interaction between proso millet quality and Baijiu flavor, and determining whether the addition of key flavor microorganisms can further enhance Baijiu quality.In addition, differences in flavor analysis results determined using GC × GC-TOF-MS and GC-IMS need to be further investigated.

Conclusion
In this study, we evaluated the changes in microbial communities and volatile flavor compounds in proso millet Baijiu, a new Baijiu product of Shanxi light-flavored Baijiu, as well as the relationship between them.Proso millet significantly changed the main flavor compound composition of light-flavored Baijiu from ethyl acetate, ethyl hexanoate, ethyl hexanoate dimer, ethyl butanoate, ethyl lactate, and butyl acetate to oct-2-ene, 2-butanol, propyl propanoate, 2-pentenal, and 4-methylpentanal.In addition, proso millet significantly changed the microbial communities in light-flavored Baijiu during fermentation, particularly during the early stages of fermentation (0-14 days).Furthermore, the formation of special flavor compounds in proso millet Baijiu was significantly correlated with the presence of fungi from the Rhizopus, Papiliotrema, Wickerhamomyces, Aspergillus, and Thermoascus genera but negatively correlated with the presence of bacteria from the Weissella, Acinetobacter, Psychrobacter, Pseudarthrobacter, Bacteroides, and Saccharopolyspora genera.The low alcohol production rate observed in Fenjiu may be due to the presence of fungi from Psathyrella genus and bacteria from the Staphylococcus, Kroppenstedtia, Brevibacterium, and Acetobacter genera, which exhibit a significantly high abundance during fermentation.Proso millet significantly changed the flavor of lightflavored Baijiu by inducing the formation of special microbial communities but did not increase alcohol concentration.This study lays a foundation for future research on improving the flavor of light-flavored Baijiu by modifying fermentation materials.
such as molecular weight, retention index (Ri), retention time (Rt), and mobility time (Dt) (Table

FIGURE 2
FIGURE 2Top 40 fungal genera (A) and species (B) obtained from the samples MA-MD and GA-GD.

FIGURE 6
FIGURE 6Gallery plot of the volatile flavor compounds based on the gas phase ion mobility spectrum.The numbers represent unidentified compounds in the mobility library.

TABLE 1 Summary of diversity indices for microbial communities based on 16S rRNA and ITS1 gene sequencing data.
All data are presented as mean ± standard deviation (n = 3).Values with different letters in a column indicate significant differences at p < 0.05 as determined by one-way ANOVA Duncan's test.

TABLE 2
Identification of flavor compounds in fermented grain samples at different stages of fermentation based on gas chromatography-ion mobility spectrometry.