Identification of Dominant Microbes and Their Successions During Solid-State Fermentation of Luzhou-Flavour Liquor Based on High-Throughput Sequencing Following Culture

Abstract

Microorganisms are crucial for the liquor brewing process and substantially impact liquor flavour and quality; therefore, understanding microbial succession is necessary. Most studies use a single-method approach and fail to provide an in-depth analysis. We aimed to combine traditional culture method with high-throughput sequencing (HTS) to identify the microbial diversity and succession in Luzhou-flavour fermentation. HTS revealed 932 bacterial and 980 fungal operational taxonomic units. 16S rDNA, 26S D1/D2 rDNA, and ITS v4/v5 isolated and identified 256 bacterial and 130 yeast strains. Population succession analysis showed that the dominant populations were yeasts, Lactobacillus, and Bacillus (early stage), and yeasts and Lactobacillus (late stage). Lactobacillus, Pichia, Bacillus, and Candida were abundant among all three layers of fermented grains. However, C. ethanolica, Saccharomycetes sp., and an unidentified Saccharomyces cerevisiae were more abundant in the lower layer than in the middle and upper layers, while L. parabuchneri, Oceanobacillus oncorhynchi, and Thermoactinomyces sp. were present only in the lower layer. Correlations among enzyme activity, volatile production, and dominant microbes during fermentation indicated that P. fermentans, L. suebicus, L. acetotolerans, P. kudriavzevii, P. exigua, and B. tequilensis were significantly affected during brewing. Our results lay a foundation for elucidating the microbial fermentation mechanism of Luzhou-flavour liquor and will assist in improving traditional liquor brewing quality and efficiency.

1. Introduction

Alcoholic beverages are popular worldwide, and the types of non-distilled and distilled alcoholic beverages vary in different countries. Traditional Chinese Baijiu (also known as Chinese spirit) is one of the oldest known distilled spirits [1,2,3]. Compared with other distilled liquors, in traditional Chinese liquors, the ethanol content is typically 40–55% [3,4]. In China, raw materials and brewing processes vary with the environment in different regions, leading to substantial variations in product flavour. Based on flavour characteristics, the traditional Chinese Baijiu flavour type can be divided into strong (Luzhou), jiang (Maotai), mild (Fen), and nine other flavours, with the Luzhou, Maotai, and Fen flavours considered to be the basic flavour types [1,2]. The brewing process of traditional Chinese Luzhou-flavour liquor involves solid-state fermentation for 2 months in an open environment with grains (fermentation substrates) and Daqu (saccharifying fermenter). The characteristic flavour of Luzhou-flavour liquor, which accounts for over 70% of the entire liquor production [5], is derived from ethyl caproate [6].

Microorganisms play a key role during the Baijiu brewing process. Their growth and metabolism have a substantial impact on the flavour and quality of Baijiu. Therefore, understanding their inter-relationship, the associated enzyme activity, and how this interaction ultimately impacts liquor quality and flavour is essential. Most studies have investigated the microbial diversity contributing to Luzhou-flavour liquor and its starter culture using only a single-method approach—traditional culture or molecular biology. For instance, Xiang et al. [7] and Li et al. [8] used small subunit rRNA profiles to evaluate microbial succession, providing a preliminary perspective for understanding the dynamic changes in the microbial community. However, this method has certain limitations in terms of the comprehensiveness and accuracy of detection. Subsequently, quantitative fluorescence polymerase chain reaction (PCR) was used to accurately quantify Streptomyces, which produces oxytetracycline during the solid-state fermentation of Baijiu. This technology can quantitatively analyse specific microorganisms but lacks the ability to analyse the overall diversity of the microbial community. Liang et al. [9] used a combination of PCR-denaturing gradient gel electrophoresis and quantitative polymerase chain reaction to analyse bacterial communities in the pit mud of Luzhou-flavour Baijiu factories in different regions. To a certain degree, this methodology overcomes the shortcomings of the single-method approach; however, it still does not provide a complete understanding of microbial diversity. Thus, the development of new approaches is essential for obtaining a more comprehensive understanding of microbial interactions affecting the flavour and quality of Baijiu liquor.

Technological developments have led to the emergence of high-throughput sequencing (HTS) technology. Fu et al. [10] determined the microbial community structure in pits of Luzhou-flavour Baijiu using metagenomic sequencing, assembly, and classification methods. This provided an understanding regarding the types, quantities, and inter-relationships of microorganisms at a broader level, thereby greatly improving our understanding of microbial communities. Liu et al. [11] used Illumina MiSeq sequencing (HTS technology) and liquid chromatography–mass spectrometry to study bacterial communities in fermented grains (zaopei). In addition to showing that the bacterial abundance and percentages of Lactobacillus and Pseudomonas increased with an increase in the depth of zaopei, they analysed the relationship between microorganisms and the fermentation process at a molecular level.

Compared with traditional technologies, HTS can sequence a large number of DNA or RNA entities in a sample, thereby comprehensively and efficiently revealing the composition, structure, and function of the microbial community [12]. HTS provides a powerful tool for in-depth research on microbial community succession, enabling a deeper understanding of the microorganisms involved in the Baijiu brewing process and elucidating additional microbial mechanisms that affect the fermentation process and liquid quality [13].

In this study, we aimed to combine traditional culture (plate separation method) and HTS technology to analyse the microbial diversity and population succession during the solid-state brewing of Luzhou-flavour liquor to establish the dominant microorganisms.

2. Materials and Methods

2.1. Materials

A GT9612 gradient thermal cycler (Hangzhou Bio-Gener Technology Co., Ltd., Hangzhou, China) was used for DNA amplification. A CF15R centrifuge (Hitachi, Tokyo, Japan) and a JK600C electrophoresis device (Beijing Junyi-Dongfang Electrophoresis Equipment Co., Ltd., Beijing, China) were used as well. An Alpha Innotech imager (San Leandro, CA, USA) was employed for gel imaging, and an Illumina sequencing platform (San Diego, CA, USA) for sequencing. Taq PCR Master Mix and primers (27f and 1492r) were purchased from Beijing Auwigene Technology, Inc. (Beijing, China). All standard laboratory reagents were purchased from Beijing Hua’aozheng Technology Co., Ltd. (Beijing, China).

Based on the analysis of 16S rRNA gene sequences, we isolated and identified bacteria in wine mash at different stages of the brewing process of Luzhou strong-flavour liquor. A homologous sequence search (BLAST2.0) was performed using the GenBank database to compare the homology of the test strains with sequences of known bacteria [14]. Subsequently, the sequence of the 26S rRNA D1/D2 region of the yeast strains isolated from the wine at different stages was analysed. The basic search tool BLAST was used for local sequence alignment to identify homologous sequences in the gene sequence data (GenBank). The test strains were compared with the corresponding sequences of known yeasts. PCR amplification of the 26S rRNA D1/D region of the isolated yeast was performed, and the amplification products were analysed using single-strand conformation polymorphism and sequence determination [15,16]. Strains with a similarity rate of >99% were defined as belonging to the same species, and those with a similarity rate of 95% to 99% were defined as belonging to the same genus. Simultaneously, the species and quantity of bacteria and yeast at different stages were summarised and collated according to the analysis and identification results. The law of bacterial populations in the process of wine mash fermentation was analysed to lay a foundation for clarifying the main microorganisms involved in the brewing process of Luzhou-flavour liquor.

2.2. Luzhou-Flavour Liquor Manufacturing Process and Sample Collection

The production process of Luzhou-flavour liquor has been reported by us previously [17]. The liquor was prepared from cereals (sorghum only or mixed with corn, rice, wheat, and glutinous rice), Daqu, rice hulls, and water in a pit (8–10 m³) using natural solid-state fermentation for 2–3 months, followed by solid-state distillation at atmospheric pressure [18]. Daqu acts as a saccharification–fermentation starter, similar to Japanese koji. Daqu was prepared as previously described [15,16,17]. The Daqu production process consists of three stages. The first stage involves the production of brick-shaped Daqu (raw wheat material). The second stage allows the microorganisms to accumulate and produce aromatic metabolites. During this process, the temperature was changed three times: to a low temperature (days 3–5), a high temperature (days 5–7), and then a low temperature once more (days 7–12). The third stage is the maturation stage, in which the temperature of Daqu decreases to the ambient temperature and further drying for storage. The preparations were stored for two to three months. For the fermentation of the Luzhou-flavour liquor, the three layers of fermented grains from the last fermentation were treated. The upper layer was distilled to obtain the liquid and then discarded. The middle layer was mixed with rice hulls, distilled for liquor, cooled to room temperature, and supplemented with 20% prepared Daqu for use as the upper layer of fermented grains in the next fermentation cycle (2–3 months). The lower layer was mixed with soaked fresh sorghum (approximately 33% water) and distilled to obtain liquor. The mixture was cooled to room temperature after distillation and supplemented with 20% Daqu for use as the middle- and lower-layer fermented grains in the next fermentation cycle (2–3 months). The procedure was repeated until the end of each fermentation period. In each cycle, fresh sorghum was added to the lower layers and maintained at a mass ratio of 2:1. The newly distilled liquor was stored in pottery at a low temperature for over a year, after which it was blended and filled into the final product.

The cell temperature increased rapidly from days 0 to 12 (peak) and gradually decreased thereafter. The upper, middle, and lower layers of the fermented grains in the pit were selected for sampling. Days 0, 4, 8, 12, 20, 30, and 44 were selected as the sampling time points, and the positions were selected based on changes in cellular temperature [19]. Samples were obtained from different layers; three samples were collected at each point and mixed for further analysis by modification.

2.3. Analysis of Glucoamylase and Liquefying Enzyme Activity and Volatile Compound Content in Grains During Fermentation

Glucoamylase and liquefying enzyme activities during fermentation were determined using methods described by Yang et al. [18]. Volatile compounds were analysed using headspace solid-phase microextraction, followed by gas chromatography–mass spectrometry (HS-SPME-GC-MS). Finally, a typical canonical correlation analysis was performed on the dominant microbes, enzymes, and volatile compounds using constrained correspondence analysis in the R vegan software package (version 1.17-3) [18,20]. The correlations between bacteria and fungi, as well as the saccharifying enzyme, liquefying enzyme, and protease activities in the fermented grains of Luzhou-flavour liquor, were analysed using constrained correspondence analysis in the R vegan software package [20].

2.4. High-Throughput Sequencing (HTS) Analysis

2.4.1. Extraction of Total DNA from Samples for HTS

High-quality total DNA was extracted according to the standard method described by Narayan et al. [21].

2.4.2. Bioinformatics Analysis

Data Quality Control

The concentration of the purified PCR products was measured using a Qubit^® 2.0 fluorescence spectrophotometer (Invitrogen, Carlsbad, CA, USA). Samples with different barcodes were mixed at equal masses to construct a database using the KAPA HyperPrep kit (KAPA Biosystems, Boston, MA, USA). Using 500 ng of DNA as the input, the sizes and molar masses of the fragments were detected using an Agilent 2100 bioanalyzer system (Santa Clara, CA, USA) to perform quantitative PCR. Qualified entries were pooled and subjected to paired-end HTS using an Illumina MiSeq PE300. The read length for each sequence was 300 bp.

The original image data of the HTS generated by the Illumina MiSeq PE300 were converted to raw sequences after base recognition and then stored in the FASTQ format. The files contained sequence information and corresponding sequencing quality information. Quality checks were performed on these files using fastqc with the following data filtering steps: removal of reads with adapters, discarding paired-end reads with a percentage N (unspecified base) > 10%, and removal of low-quality reads (criteria: Q20 ≥ 90%).

Read Merging and Optimisation

Reads were joined into tags using connecting overlapping pair-end (COPE, V1.2.3.3) software according to the overlaps among paired-end reads. The barcodes on both sides of the tags and primer sequences were trimmed using a code developed by our group to obtain raw tags; tags < 200 bp in length and chimeras were filtered to obtain clean tags. To ensure the quality of data for subsequent analysis, the raw tags from paired-end read merging were optimised to obtain clean tags using Mothur software (version 1.48.1) [22]. Raw tags were screened for anomalies. Unique tags were selected to minimise redundancy and were compared with the SILVA reference database (version 119, http://www.arb-silva.de/, accessed on 15 July 2020) using the Needleman–Wunsch algorithm [23] for target region matching. These clustered low- and high-abundance tags had minor mismatches to reduce erroneous operational taxonomic unit (OTU) determination. Chimeric sequences were identified using the UCHIME algorithm (v4.2, http://drive5.com/uchime/, accessed on 20 July 2020) [24]. Sequences annotated as nonbacterial or fungal were excluded. Bioinformatics parameters for chimera removal used UCHIME algorithm settings, and OTU clustering identity used OTUs with 97% similarity.

3. Results

3.1. HTS Rarefaction Curves

Rarefaction curves were plotted for the species present in a number of randomly selected samples together with their OTU numbers. These can be used to compare species richness in samples with different amounts of sequenced data and to determine whether the amount of sequenced data is reasonable. A flat curve indicates a reasonable amount of sequenced data—that is, additional sequencing will produce only a small number of new OTUs—whereas a steeper curve indicates that additional new OTUs can be produced by further sequencing [25]. Thus, the rarefaction curves reflect the sequencing depth of the sample. The relationship between the number of sequences obtained during liquor brewing and fungal diversity is shown in Figure 1A. The curves approached a plateau when 25,000 sequences were randomly selected, indicating sequencing saturation and good species coverage. The 1,045,836 sequences obtained from the ITS1 region of the fermented grain samples (an average of 49,801 effective sequences per sample) were representative of winemaking fungal populations. The species richness of the upper layer of fermented grains was the highest on the 44th day and the lowest on the 8th day (Figure 1A), whereas that of both the middle- and lower-layer fermented grains was the highest on the 12th day and the lowest on the 0th day (Figure 1B,C). The plots indicated satisfactory sequencing results for the samples and provided a general understanding of the differences in species diversity between the upper, middle, and lower fermented grain layers.

Figure 1a–c (bottom) show the relationship between the number of sequences obtained during liquor brewing and the bacterial diversity. The curves approached a plateau when 15,000 sequences were randomly selected, indicating sequencing saturation and good species coverage. The 976,670 sequences obtained for the 16S V3/V4 regions (an average of 46,508 effective sequences per sample) are representative of winemaking bacterial populations. The OUT number ranking was lower layer > middle layer > upper layer on the 0th day under the same sequencing depth, and it was almost the same for all three layers on the 8th day. The OUT number ranking was upper layer > middle layer on the 8th day, lower layer > upper layer > middle layer on the 12th day, and upper layer > middle layer > lower layer on the 20th day. The number was small and similar for all three layers on the 30th day and increased on the 44th day, but it still did not vary markedly across the layers. For the upper layer, the bacterial diversity increased during days 0–8 and gradually decreased during days 8–12. During days 12–44, it increased and decreased, followed by another small increase. During the elaboration of Luzhou-flavour liquor, the upper layer becomes the middle layer and eventually the lower layer in subsequent fermentation cycles, with the microorganisms therein supplying and controlling the microbial populations in the other layers.

3.2. HTS Shannon–Wiener Curve

The Shannon–Wiener curve reflects the sample microbial diversity. Microbial diversity indices of samples sequenced at different depths were plotted to reveal the microbial diversity as different quantities of data were sequenced. A flat curve indicates a sufficient amount of sequencing data containing most of the microbial information in the sample. The Shannon index on the vertical axis reflects the diversity of the sample species [26] in the Shannon–Wiener curves for fungi (Supplementary Figure S1A–C) and bacteria (Supplementary Figure S1a–c) in the upper, middle, and lower layers of the fermented grains. The point where all curves approach a plateau indicates that sufficient data have been sequenced and the results are illustrative of microbial populations. The figures clearly show the differences in fungal and bacterial diversity of the layers at different brewing times.

3.3. Dominant Microbes in the Fermented Grain Layers Based on HTS

Figure 2A–C shows the results of HTS analysis at the genus level. Figure 2a–c shows the results of the dominant fungi in the upper, middle, and lower layers based on the number of sequences of each OTU, which were compared with the model strains in GenBank and the isolated strain data in the Baijiu data bank. It can be seen that the fungi that occupied > 90% of the microbial population in all three layers comprised Pichia fermentans, Pichia exigua, Pichia kudriavzevii, Candida ethanolica, Candida humilis, Aspergillus chevalieri, Saccharomycetes, Geotrichum candidum, Vishniacozyma victoriae, Scytalidium thermophilum, Saccharomyces cerevisiae, Cladosporium phlei, and Vishniacozyma tephrensis. Their layer distributions followed the same pattern with only small differences. For example, P. fermentans was proportionally the highest in the three layers, with its abundance decreasing as the deeper layers were sampled. Conversely, the abundances of P. exigua and P. kudriavzevii increased in the deeper layers. These dominant strains act as important references for studying functional microorganisms in Baijiu brewing and flavour formation in the future.

Figure 3A–C shows the results of the HTS analysis at the genus level. Figure 3a–c shows the results of the dominant bacteria in the upper, middle, and lower layers based on the number of sequences of each OTU, which were compared with the model strains in GenBank and the isolated strain data in the Baijiu data bank. The following dominant bacterial species were identified (Figure 3): L. suebicus, Bacillus subtilis, Acetobacter pasteurianus, L. acetotolerans, L. brevis, Staphylococcus gallinarum, L. suebicus, Escherichia coli, Pantoea agglomerans, Pediococcus pentosaceus, Weissella cibaria, Enterobacter cloacae, B. megaterium, and L. plantarum. Conversely, B. cereus, L. porcinae, Thermoactinomyces sanguinis, Halomonas stevensii, and L. farciminis were dominant species that appeared only in two layers. B. coagulans (upper panel), L. curvatus (middle panel), L. parabuchneri, Oceanobacillus oncorhynchi, and T. sanguinis. (lower panel) were dominant in only one layer. In particular, L. suebicus was the most abundant species in all three layers and can, thus, be considered the dominant bacterium in the brewing of this liquor. Different species and their proportions may result in variations in the fermentation of different layers.

3.4. Tree-Structure Analysis of Multiple Sample Similarity Based on the UniFrac of HTS

Visualisations such as evolution trees were constructed using the unweighted pair-group method with arithmetic means for hierarchical clustering to observe similarities and differences in microorganism evolution in different environmental samples [27]. Cluster analysis based on sample similarity was performed on the fungal communities of the upper-, middle-, and lower-layer fermented grains at various fermentation stages. The upper layer samples on day 44 differed markedly, whereas the samples on day 20 were the most similar. The day 44 middle layer samples differed the most, whereas the samples on days 8 and 20 were most similar to each other and to the day 30 samples (Supplementary Figure S2A–C). The changes were mainly detected in the day 4 and day 8 samples. The lower layer resembled the middle layer, differing only on day 44. Differences between the samples during fermentation may be related to changes in the environment and microbial populations in the layers. Cluster analysis based on sample similarity was performed on the bacterial communities of the upper, middle, and lower layers of fermented grains at various fermentation stages. Samples from all three layers were clustered into two groups, with day 12 as the division line (Supplementary Figure S2a–c, bottom). The cluster results of the upper and middle layers group the day 12 sample with those of days 0–8, whereas the cluster results of the lower layer group the day 12 sample with those of days 20–44. The lower and middle- and upper-layer samples were slightly different, whereas the upper and middle layers were more similar. These differences may be related to the changes in the environment and microbial populations in the layers.

3.5. Principal Component Analysis (PCA)

PCA was performed on the fungal and bacterial communities in the fermented grains during fermentation (Supplementary Figure S3). Fungi from the upper layer were considerably different from those from other samples (middle and lower) on day 44. These results are similar to those obtained using the cluster analysis mentioned above. At other time points, the differences were mainly apparent in the longitudinal direction of the sample, indicating the effects of the main acceptor components in PC2. Overall, the results obtained using principal component and cluster analyses for the three different layers agreed with each other. In addition, the three layers were highly similar, indicating that the results of the PCA and clustering analyses were consistent among the samples.

3.6. Population Succession of Dominant Microorganisms During Liquor Fermentation Based on HTS

Figure 4 shows a thermographic clustering map of the dominant fungi and bacteria at different fermentation stages obtained using HTS. Variations in the dominant fungi and bacteria by layer were clearly discernible. A similarity was observed in the variation of microbial composition in the layers where P. fermentans, P. exigua, C. humilis, P. kudriavzevii, A. chevalieri, S. cerevisiae, and C. ethanolica were the most prominent. P. fermentans was found to be abundant throughout the fermentation process. The levels of P. exigua, C. humilis, and S. cerevisiae initially increased and then decreased as fermentation proceeded. The numbers of P. kudriavzevii and A. chevalieri increased during fermentation. Although many microbial populations varied similarly across all three layers, some differed from each other. C. ethanolica and an unidentified Saccharomycetes, for example, were abundant in the upper and middle layers at the end of fermentation (day 44), but not in the lower layer. According to the thermographic clustering map of dominant fungi, the population succession of S. cerevisiae differed in each layer during fermentation. Thus, fungal succession in the upper and middle layers followed similar patterns, which were different from that in the lower layer. This may be related to the different environments in each layer. During Luzhou-flavour liquor solid-state fermentation, water and microbial metabolites begin to concentrate in the lower layer, producing higher-quality liquor. Upon metabolite accumulation, the acidity of the lower layer and the alcohol content slowly increase. Therefore, certain low-tolerance microorganisms die, resulting in differences between the lower layer and the middle/upper layers.

Variations in the dominant bacteria by layer were also evident (Figure 4). The dominant bacteria present in all three layers during fermentation days 0–12 were B. subtilis, A. pasteurianus, L. brevis, S. gallinarum, L. suebicus, E. coli, P. agglomerans, P. pentosaceus, W. cibaria, E. cloacae, B. megaterium, and L. plantarum, the numbers of which decreased from days 0 to 8 (as the highest fermentation temperature was reached on day 12). The dominant bacteria on days 12–44 were L. suebicus and L. acetotolerans, the numbers of which increased during the late fermentation stage. The differences in the population succession of the upper-, middle-, and lower-layer bacteria were mainly reflected in L. brevis, L. porcinae, B. cereus, T. sanguinis, L. curvatus, L. parabuchneri, O. oncorhynchi, and Thermoactinomyces sp.

The bacterial isolation and culturing efforts [14] revealed a small number of bacteria. However, it is noteworthy that the species richness was high from day 0. Among them, the number of Acetobacter malorum, Acetobacter cerevisiae, and B. methylotrophicus was slightly higher, and on day 4, the number of species of bacteria increased significantly; this increase was mainly driven by B. methylotrophicus, Acetobacter aceti, Acetobacter estunensis, Acetobacter pasteurianus subsp. pasteurianus, and B. vanillea. On day 8, the abundance and diversity of bacteria were still very high, and the fermentation temperature of the wine lees also increased; the dominant bacteria in this period were B. amyloliquefaciens subsp. plantar, B. methylotrophicus, and B. atrophaeus. On approximately day 12, the fermentation temperature of the wine lees reached its peak, with the abundance and diversity of bacteria also peaking at this point. The dominant strains in this period were B. subtilis subsp. inaquosorum, B. amyloliquefaciens subsp. plantarum, L. buchneri, L. paracasei subsp. tolerans, and Bacillus vanillea. On day 20, the abundance and diversity of bacteria began to decrease; this decrease was suspected to be driven by the observed temperature increases. On day 3, the bacteria increased, with this trend mainly driven by B. subtilis subsp. inaquosorum, B. sonorensis, B. amyloliquefaciens subsp. plantarum, B. methylotrophicus, B. vanillea, and B. atrophaeus. On day 44, the fermentation ended, and the abundance and diversity of bacteria began to decline; the dominant species at this point were B. amyloliquefaciens subsp. plantarum, B. methylotrophicus, and Gluconobacter cerinus.

3.7. Correlations Between Dominant Strains, Enzyme Activity, and Volatile Component Content

Figure 5 shows the dynamics of glucoamylase (left) and liquefying enzyme (right) activities in fermented grains in upper (A), middle (B), and lower (C) layers during Luzhou-flavour liquor production. The liquefaction power of the upper layer of fermented grains increased rapidly from days 0 to 4, then gradually decreased. A trend of first rising and then decreasing appeared after the 12th day. The liquefaction power of the middle layer was generally low and only reached a higher level at approximately day 30. The liquefaction power of the lower layer of wine lees fluctuated greatly, with significant increases in liquefaction power on the 8th and 30th days and significant downward trends on the 20th and 44th days. The difference in liquefaction enzyme activity among the three levels of the wine lees was also relatively large. Many factors affect the change of liquefaction enzyme activity, such as the community structure of microorganisms, water content, reducing sugar content, and acidity. Furthermore, the difference in factors among the levels is very large, which also causes the inconsistency in the enzyme activity changes among the layers.

Figure 6A (left) shows that the strains that correlated highly with glucoamylase activity in the upper layer were X2 (B. subtilis/tequilensis), X10 (Weissella cibaria/confuse), X8 (Pantoea agglomerans), X6 (Staphylococcus gallinarum), and X11 (Pediococcus pentosaceus), and the liquefying enzyme-related strains X4, X12 (Escherichia coli), and X13. The strains that correlated with glucoamylase activity in the middle layer (Figure 6B (left and right)) were X8, X2, X10, X6, and X4, while the liquefying enzyme-related strains were X13 and X20. The strains that correlated with glucoamylase activity (Figure 6C (left and right)) in the lower layer were X8, X2, X10, and X12, whereas the liquefying enzyme-related strains were X30, X28, and X20. The common strains that were highly correlated with glucoamylase activity were Bacillus and Weiss, whereas those related to liquefying enzymes were Bacillus and Candida.

Volatile substances in different samples from each layer were analysed using HS-SPME-GC-MS (Supplemental Tables S1–S3). Overall, 74 types of volatile substances were identified in the upper layer: 33 esters, 17 alcohols, 4 acids, 5 phenols, 5 aldehydes, 4 ketones, and 6 others; 96 types of volatile substances were identified in the middle layer: 34 esters, 18 alcohols, 4 acids, 7 phenols, 12 aldehydes, ketones, and 21 others; and 83 types of volatile substances were identified in the lower layer: 34 esters, 15 alcohols, 10 acids, 6 phenols, 8 aldehydes, ketones, and 10 others. The correlation between these strains and esters was stronger than that with alcohols, followed by that with acids. High middle-layer correlations between acids, alcohols, and strains X1 (L. suebicus) and X14 (Pichia exigua) are shown in Figure 6B (right). Higher correlations between esters and X12 (L. suebicus) were observed. Figure 6C (right) (bottom) shows strong correlations between the acids, esters, and X3 (L. acetotolerans). A strong correlation between alcohol consumption and X12 (L. suebicus) was also observed. The microbes that correlate with enzymes and volatile substances may be the functional brewing microorganisms of Luzhou-flavour liquor.

4. Discussion

Evaluation of fermented grain microbial communities may provide a fundamental understanding of the species responsible for liquor quality, in addition to generating important resources for regulating annual fermentation processes and developing strategies to ensure product quality [5]. In this study, numerous dominant bacteria and fungi were identified using a combination of HTS and culture-dependent methods, accounting for 92.5% of the total bacteria and 92.7% of the dominant fungi.

Lactococcus, Sporolactobacillus, and Streptococcus were isolated using the culture-dependent method [28]; Pichia, Saccharomycopsis, and Talaromyces were dominant on the first day, whereas the fungi Eurotium and the bacteria Burkholderia, Lactobacillus, and Streptococcus were abundant on the 7th day, and only Burkholderia and Lactobacillus were detected in the small subunit rRNA profiles on days 1, 7, and 60 [7]. Consistent with the present findings, Zhang et al. [29] showed that the genera Saccharomyces and Saccharomycopsis were the dominant fungi in Luzhou-flavour liquor. However, their observations revealed changes in the microbial population during fermentation only at the genus level and did not reveal the succession of dominant species populations, as shown in this study. Chen et al. [30] isolated the genera Candida, Pichia, and Aspergillus from fermented grains, as also observed in this study. The studies may differ not only in the method but also because the growth and metabolism of microorganisms are influenced by the substrates, water, pH, temperature, oxygen, and a complex micro-ecosystem, achieving dynamic balance through coordination and inhibition.

In a previous study, we isolated and identified bacterial microorganisms from fermented grains at different stages of Luzhou-flavour liquor by 16S rRNA gene sequence analysis; a total of 258 strains of bacteria were isolated, and the dominant bacterial microorganisms in different fermentation stages of wine lees fermentation were quickly and accurately found [14]. The bacterial diversity in the early stages of fermentation was very rich, and there were no obvious differences between the various genera. In the middle stages of fermentation, the bacteria were mainly Bacillus and Lactobacillus, whereas in the later stages of fermentation, Bacillus was the predominant genus. The results of this study show that Bacillus and Lactobacillus are the main functional microorganisms in the fermentation process of Luzhou-flavour liquor, which play an important role in the flavour formation; this also verifies the relevant research of our predecessors. For example, the microbes in the lower layer may enhance caproic acid production [31]. Yeasts were also isolated from fermented grains at different stages of fermentation, and variations among different yeast species were analysed using single-strand conformation polymorphism. Furthermore, D1/D2-region sequences of the 26S rRNA gene of 130 yeast strains belonging to 9 genera and 15 species (P. fermentans, Naumovozyma castellii, Torulaspora delbrueckii, Saccharomyces cerevisiae, P. membranifaciens, Candida humilis, Kazachstania exigua, S. fibuligera, Millerozyma farinosa, C. cabralensis, P. kudriavzevii, C. ethanolica, P. occidentalis, Zygosaccharomyces bailii, and C. rugopelliculosa) were analysed and identified.

In this study, we observed that the dominant fungal population succession was nearly identical across the layers during fermentation. In particular, the abundance of Candida ethanolica, Saccharomycetes sp., and Saccharomyces cerevisiae was higher in the lower layer than in the middle and upper layers. By convention in liquor making, liquor distilled from the lower layer is typically of higher quality than that distilled from the middle and upper layers, which may be explained by the different microorganism successions in the different layers. To date, few studies have examined Candida ethanolica. It differs from other species in the genus in that it does not assimilate nitrate, produce urease, or ferment sugars. Bacteria, which are the main driving forces in solid-state fermentation and acid formation and important for yeast growth and flavour development in various types of traditional Chinese liquors, are indispensable for Daqu preparation [17], grain fermentation [18], and pit mud [32]. Here, the middle- and lower-layer bacterial diversities were substantially higher than those in the upper layer, with the dominant bacteria common to the middle layers exhibiting similar changes. Interspecific association analysis identified 30 bacterial species. Phylogenetic analysis using 16S rDNA of the genus Bacillus isolated from the fermented grains of Luzhou-flavour liquor demonstrated that the genus is diverse, without a clearly dominant species. Wang et al. [33] studied the diversity of culturable bacteria in the Luzhou-flavour liquor brewing of Yibin and observed that the genus Bacillus was the dominant bacterium in fermentation, whereas the genera Streptomyces and Staphylococcus were secondary. These studies are limited because they were based on culturable bacteria and did not consider non-culturable microorganisms in their analyses. Zhang et al. [29] used PCR to identify bacterial communities and identified Lactobacillus as the dominant genus, as shown in the present study. However, they only analysed middle layer samples, whereas this study examined the bacteria present in all three layers during brewing, thereby allowing for a more accurate determination of the dominant species and their population succession. The differences in bacterial species between the lower and upper or middle layers were evidenced by the dominance of L. parabuchneri, O. oncorhynchi, and Thermoactinomyces sp. only in the lower layer. To observe the diversity and abundance of microorganisms present in each of the three layers more clearly, the different species and numbers of isolates obtained from each layer should be determined. L. parabuchneri is a heterofermentative lactic acid bacterium capable of reducing lactic acid content and increasing acetic acid production [31]. Excess lactic acid inhibits the growth and proliferation of beneficial microorganisms in Luzhou-flavour liquor brewing, such as caproic acid-producing bacteria, which have limited acid tolerance and only live at weakly acidic or nearly neutral pH. As a high 708 lactic acid content impairs their vitality, L. parabuchneri is important for the suppression of lactic acid and elevation of caproic acid-producing bacteria. Studies on O. oncorhynchi, a salt-tolerant, Gram-positive, rod-shaped, flagellate, obligate alkaliphile that produces oval spores [34], are limited. He et al. [35] characterised Actinobacteria isolated from Luzhou-flavour liquor and studied the fragrances of their metabolites, resulting in the separation of 123 strains producing volatile products. Their metabolites include numerous acids, alcohols, and esters that constitute their liquor flavour. He et al. [35] also observed that Streptomyces mutabilis, S. vinaceusdrappus, S. coelicoflavus, and S. violascens were involved in increasing the content of flavour compounds, such as ethyl acetate, ethyl caproate, ethyl lactate, butanoic acid, and furfural. The different characteristics of these bacteria may affect fermentation during Luzhou-flavour liquor production. The correlation between enzyme activity and changes in volatile compound content was similar to that previously reported for stabilising persistent microbial succession. This indicated that P. fermentans, L. suebicus, L. acetotolerans, P. kudriavzevii, P. exigua, and B. tequilensis in fermented grains significantly affected the brewing of Luzhou-flavour liquor. The succession of bacterial communities in fermented grains was analysed using HTS technology, and the correlation between the alternation of communities at different fermentation stages and variation of environmental factors was analysed using the Mantel test [36]. The results showed that there were 397 genera of microorganisms in the fermentation process of fermented grains, and Lactobacillus, Bacillus, Weissella, Dysgonomonas, Comamonas, and Ruminococcaceae were the dominant genera (relative abundance > 1.0%). However, in the present study, the change rule of microorganisms in the upper, middle, and lower three levels of the mash was very similar. The number of bacteria was relatively high on days 0–12, comprising mainly Lactobacillus, Acetobacter, and Bacillus, while the fungi were at a slow growth stage except for the rapid growth of P. fermentans. From 12 to 44 days, most of the bacteria disappeared, leaving only Lactobacillus, and L. acetotolerans increased significantly, while the fungi were mainly represented by large numbers of some yeasts. The abundance of P. fermentans was still high, but it was unclear whether the numbers included dead strains from the previous period; P. exigua and P. kudriavzevii also showed a significant increase. Alternatively, samples from each period of the fermentation cycle of the fermented grains were isolated and identified, with 258 strains of bacteria belonging to 30 species and 130 strains of fungi belonging to 15 species. By comparing these with the results of HTS analysis, the dominant microorganisms were determined to be B. subtilis subsp. inaquosorum; B. methylotrophicus; A. oxydans; A. cerevisiae; A. malorum; L. paracasei subsp. tolerans; B. amyloliquefaciens subsp. plantarum; A. aceti; L. brevis; A. pasteurianus subsp. pasteurianus; B. aryabhattai; Pseudomonas veronii; A. punensis; B. vanillea; L. pentosus; A. pomorum; A. sicerae; B. atrophaeus; L. buchneri; B. flexneri; L. hilgardii; B. sonorensis; L. parafarraginis; Staphylococcus epidermidis; L. xiangfangensis; L. plantarum subsp. plantarum; and L. acetotolerans, accounting for 92.5% of the total bacteria in the high-throughput group. The dominant fungi were C. rugopelliculosa; P. fermentans; S. cerevisiae; P. membranifaciens; C. humilis; S. fibuligera; C. cabralensis; P. kudriavzevii; C. ethanolica; and P. occidentalis, accounting for 92.7% of the high-throughput ITS data, indicating that the isolation and identification results were ideal and the experimental purpose was achieved.

Therefore, HTS analysis can accurately reveal the population composition, structure, and function of microorganisms, as well as their succession. It can also identify dominant microorganisms, thereby providing a foundation for further research.

Traditional Chinese Luzhou-flavour liquor has a long history, spanning thousands of years. However, its traditional production technology has gaps compared to other brewing technologies, mainly because functional microorganisms have not been clearly identified to date; therefore, the specific brewing mechanism remains unknown [17]. Many studies have been conducted to understand this mechanism [5,18,37,38]. It was recently reported that the dominant bacteria in fermented grains of Luzhou-flavour liquor were Lactobacillus, Bacillus, Weissella, Dysgonomonas, Comamonas, and Ruminococcaceae [36], and the dominant fungi in the middle layer of fermented grains were mainly Kazachstania, Aspergillus, Thermomyces, Thermoascus, and Eurotium [39]. However, the results of the present study showed that the dominant populations at the early fermentation stage were Lactobacillus, Bacillus, and yeasts, and Lactobacillus in the late stage of fermentation. These differences may be related to the differences in the samples obtained using different brewing technologies [5,18]. These observations will help elucidate the functional microorganisms and mechanisms underlying traditional Chinese Baijiu production and, therefore, will have a considerable impact on improving the quality and efficiency of traditional liquor brewing. However, to confirm their functionality, samples from each period are required for transcriptome analysis. Finally, pure culture experiments are required to confirm the microbes associated with flavour.

5. Conclusions

To explore the mechanism underlying the microbial interactions involved in the Luzhou wine brewing process, the changes in the microbial community of fermented grains were tracked using HTS and a traditional culture separation technology. The dominant microorganisms in the early and late stages of fermentation were yeast, lactic acid bacteria, and spore-forming bacteria and yeast and lactic acid bacteria, respectively. In the three levels of fermented grains, the content of alcohol substances was the highest, with the middle and lower levels being higher than the upper level, and with the lower-level acid substances being higher than those in the middle level. Finally, regarding the correlation of the phase of bacteria and volatile substances in the fermentation process, the microorganisms related to the volatile substances in the upper, middle, and lower levels were consistent. Our results lay a foundation for elucidating the microbial fermentation mechanism of Luzhou-flavour liquor and will assist in improving traditional liquor brewing quality and efficiency.