Abstract

   In this study, a fortified Daqu (FF Daqu) was prepared using high
   cellulase-producing Bacillus subtilis, and the effects of in situ
   fortification on the physicochemical properties, flavor, active
   microbial community and metabolism of Daqu were analyzed. The
   saccharification power, liquefaction power, and cellulase activity of
   the FF Daqu were significantly increased compared with that of the
   traditional Daqu (CT Daqu). The overall differences in flavor
   components and their contents were not significant, but the higher
   alcohols were lower in FF Daqu. The relative abundance of dominant
   active species in FF Daqu was 85.08% of the total active microbiota
   higher than 63.42% in CT Daqu, and the biomarkers were Paecilomyces
   variotii and Aspergillus cristatus, respectively. The enzymes related
   to starch and sucrose metabolic pathways were up-regulated and
   expressed in FF Daqu. In the laboratory level simulation of baijiu
   brewing, the yield of baijiu was increased by 3.36% using FF Daqu.

   Keywords: Daqu, Cellulase, Active microorganisms, Carbohydrate enzymes,
   Physical and chemical indicators, Flavor substances, Non-volatile
   metabolites

Highlights

     * •
       Inoculating high-yield cellulase Bacillus subtilis to make
       fortified Daqu.
     * •
       Bioturbation of functional strains improved the quality of Daqu.
     * •
       Bioturbation affected the microflora, enzyme spectrum and metabolic
       of Daqu.

1. Introduction

   The quality of Daqu plays a key role in the formation of the body and
   flavor of baijiu ([41]Yang, Fan, & Xu, 2021) Daqu provides the driving
   force for fermentation, provides a large number of functional
   microorganisms and complex enzymes in baijiu-making, and uses raw
   materials for biochemical reactions such as alcoholic fermentation and
   aroma production, thereby giving baijiu a unique flavor ([42]He, Huang,
   Wu, Jin, & Zhou, 2020). In addition, Daqu after fermentation and
   storage can provide a fermentation matrix for the brewing system as a
   raw material for brewing, and its own rich flavor substances and its
   precursors can also directly or indirectly enter the baijiu body during
   the brewing process ([43]Xu et al., 2017). Therefore, it is
   particularly important to improve the microbial community structure,
   physical and chemical properties and the flavor substances and flavor
   precursors of Daqu to improve the baijiu yield and quality.

   In the process of Daqu open fermentation, natural inoculation and
   entrapment of natural microorganisms rely on the accumulation of
   endogenous bioheat of microorganisms to promote microbial interaction
   and drive functional microbial community succession, so as to form a
   stable Daqu microecology ([44]Liu et al., 2023; [45]Yang et al., 2021).
   The interaction between microorganisms weaves a cooperative network of
   species working together for common goals for the evolution of Daqu
   microecosystem ([46]He et al., 2020). However, the microbial succession
   of non-artificially controlled evolution limits the occurrence and
   effectiveness of synergistic effects between microbial communities, and
   it is difficult to ensure the unity of mass and functional expression
   of Daqu ([47]Li et al., 2020). In recent years, the enhancement of
   functional microorganisms has been gradually applied to traditional
   fermented foods, such as baijiu, vinegar, soy sauce and kimchi, and the
   biodisturbance effect based on microbial enhancement has significantly
   changed the taxonomic composition, structure and interspecific
   interaction of microbial communities in traditional fermentation
   systems and the expression of enzyme-coding genes, which is conducive
   to the improvement of product quality and yield ([48]He et al., 2020;
   [49]He, Huang, Zhou, Wu, & Jin, 2019; [50]Li et al., 2020). Therefore,
   targeted use of functional strains is one of the methods to improve the
   quality of Daqu.

   Studies have shown that cellulase has certain activity in Daqu
   fermentation and fermented grains fermentation, and cooperates with
   amylase, glucoamylase, esterase and various microorganisms to ensure
   the smooth start and progress of fermentation ([51]Shi et al., 2022).
   The use of cellulase can significantly improve the utilization rate of
   raw materials, reduce the fusel oil content and improve the aroma
   substances in baijiu ([52]Zhuansun et al., 2022). In order to ensure
   the safety of baijiu, exogenous non-fermenting substances are
   prohibited from being added to the fermentation process, and functional
   microorganisms can be used as carriers to enter the fermentation system
   of fermented grains for functional expression ([53]Li, Liu, Liu, Hui, &
   Pan, 2023). In addition, Bacillus bacteria are one of the important
   functional flora in the microecology of baijiu-brewing, which can
   secrete a variety of enzymes that hydrolyze raw material components,
   promote the smooth progress of fermentation process and the full use of
   biomass ([54]Li, Fernandez, Vederas, & Gänzle, 2023; [55]Tang et al.,
   2023; [56]Tong et al., 2022). However, due to the little understanding
   of the biological disturbance mechanism of functional strain
   enhancement in Daqu, the progress of the enhancement application of
   functional microorganisms in Daqu is greatly limited.

   At present, the study on the use of high-cellulase strains for the
   fortification of Daqu has not been reported, and there is less
   understanding of the differences in the active microbiota of Daqu
   before and after fortification. To this end, this study used
   metatranscriptomics, metabolomics, physicochemical factors and enzyme
   activity determination and HS-SPME-GC–MS to analyze the bioturbation
   effect of high-yield cellulase Bacillus subtilis on Daqu.

2. Materials and methods

2.1. Production and sampling of Daqu

   The high-yielding cellulase Bacillus subtilis was screened from
   fermented grains, and this strain was screened from our previous work
   ([57]Liu et al., 2020). Fortified Daqu (experimental group FF Daqu) and
   traditional Daqu (control group CT Daqu) were produced from Daqu-making
   workshop of Henan Yangshao Distillery Co., Ltd., and the fortification
   of functional strains was based on the production process of
   traditional Daqu, which was carried out by uniform spraying of
   bacterial liquid onto the surface of the pressed and molded Daqu, with
   the concentration of bacterial liquid of 10^6 cells/mL (Fig. S1A).
   Mature Daqu was collected from the upper, middle and lower layers of
   the stack and the left and right sides of the middle layer (Fig. S1B).
   Daqu was broken from the middle, the section was observed, the aroma
   was smelled and the quality was judged (Fig. S1C). Crush each qualified
   daqu and mix evenly. The samples are packed and sealed in a sterile
   sampling bag numbered FF-1, FF-2, FF-3, FF-4, FF-5, CT-1, CT-2, CT-3,
   CT-4, CT-5. Transport to laboratory on dry ice and store at −80 °C.
   Among them, samples numbered 1, 2, and 3 were used for physicochemical
   index determination, volatile compound analysis, and metatranscriptome
   extraction. All samples were analyzed for non-volatile metabolites.

2.2. Analysis of physical and chemical factors and enzyme active substances
of Daqu

   The moisture content of Daqu was determined by oven-drying constant
   weight method, that is, dried at 101–105 °C, and the weight change
   between the last two measurements was <0.002 g. The total titration
   acidity was determined by acid-base neutralization method (In short,
   0.1 mol/L NaOH was used to titrate the sample to pH 8.2, so as to
   calculate the acidity of Daqu). Starch was hydrolyzed by hydrochloric
   acid, and the amount of glucose was determined by Fehling method to
   reflect starch content ([58]Liu, Li, et al., 2023). The fermentation
   power was determined by carbon dioxide weight loss method (the weight
   of carbon dioxide produced by liquid fermentation at 35 °C for 72 h
   with sorghum juice as starter). Under the condition of 35 °C, pH 4.6,
   the time spent on the disappearance of the blue-violet reaction of the
   Daqu hydrolysis soluble starch solution to iodine was used to reflect
   the liquefaction power. The saccharification power was determined by
   Fehling method, the esterification power was determined by
   saponification method, and the amino acid nitrogen was determined by
   potentiometric titration method ([59]Yang, Fan, & Xu, 2021; [60]Yang,
   Wang, et al., 2021). Cellulase enzyme activity was measured using the
   CMCA-DNS method according to the standard QB 2583–2003 (Cellulases),
   which was determined by the quantity of reducing sugar that was
   hydrolyzed from the substrate sodium carboxymethyl cellulose by 0.5 mL
   of Daqu filtrate in 30 min at 50 °C (U, mg/g·min) ([61]Shi et al.,
   2022). The linear equation of glucose obtained in this experiment was
   y = 0.1342× - 0.1619，R^2 = 0.991.

2.3. Analysis of volatile compounds in Daqu

   Sample preparation: 1 g of Daqu was taken into a headspace bottle, 2 g
   of NaCl, 5 mL of distilled water were added, the stopper was screwed
   tightly and shaken well. The headspace vial was preheated in a water
   bath at 60 °C for 10 min, and a solid-phase CAR/PDMS (75 μm CAR/PDMS,
   carbon molecular sieve/polydimethylsilane) extraction fibre tip was
   inserted into the silicone stopper of the headspace vial, and was
   inserted into the sample for headspace adsorption for 30 min. GC
   conditions: the chromatographic column was HP-FFAP
   (30 m × 0.32 mm × 0.25 μm); no shunt, flow rate 1.21 mL/min; inlet
   temperature: 250 °C; heating procedure: 40 °C was held for 3 min,
   5 °C/min to 60 °C without holding, 8 °C/min to 230 °C and held for
   7 min; mass spectrometry (MS) conditions: the interface temperature was
   220 °C, and the ionization mode was an electron ionization source with
   an electron energy of 70 eV, and the ion source temperature was 200 °C
   ([62]Kang, Chen, Han, & Xue, 2022). The compounds were quantified by
   calculating the percentages of GC peak areas. Then analysis of relative
   odor activity value (ROAV) based on methods reported in the literature
   ([63]Mu et al., 2023).

2.4. Metatranscriptomic analysis of Daqu

2.4.1. RNA extraction and shotgun sequencing

   Total RNA was extracted using the RNA Power Soil® Total RNA Isolation
   Kit (12866–25) (MoBio, USA) according to the manufacturer's
   instructions. The extracted RNA was completed using 1.5% agarose gel
   electrophoresis and UV spectrophotometer for quality determination and
   quantification. Meanwhile, RNA integrity (RIN ≥ 5.5) was detected using
   an Agilent 2100 (Agilent, USA). After removal of ribosomal RNA for
   microbial RNA, Illumina's TruSeq Stranded mRNA LT Sample Prep Kit
   (Illuminia, USA) was used for reverse transcription as well as
   macro-transcriptome birdshot sequencing library construction. Each
   library was sequenced by Illumina NovaSeq platform (Illumina, USA) with
   PE150 strategy at Personal Biotechnology Co., Ltd. (Shanghai, China).

2.4.2. Metatranscriptomics analysis

   Raw sequencing reads were processed to obtain quality-filtered reads
   for further analysis. First, sequencing adapters were removed from
   sequencing reads using Cutadapt (v1.17). Secondly, low quality reads
   were trimmed using a sliding-window algorithm in fastp (v0.20.0). Then,
   Ribosomal RNA was removed by SortMeRNA (v4.2.0) using its default rRNA
   reference database.

   Once quality-filtered reads were obtained, taxonomical classifications
   of metatranscriptomics sequencing reads from each sample were performed
   using Kaiju with greedy-5 mode against an nr-derived database, which
   included proteins from archaea, bacteria, viruses, fungi and microbial
   eukaryotes. All raw sequences were deposited in the NCBI Sequence Read
   Archive under accession number PRJNA1087545.

2.5. Analysis of non-volatile compounds

   Sample preparation: 1). Accurately weigh an appropriate amount of
   sample into a 2 mL centrifuge tube, add 100 mg glass bead. 2). Add
   1000 μL 50% methanol water (stored at 4 °C), vortex for 30 s. 3). Put
   the centrifuge tube containing the sample into the 2 mL adapter matched
   with the instrument, immerse it in liquid nitrogen for rapid freezing
   for 5 min, take out the centrifuge tube and thaw at room temperature,
   put the centrifuge tube into the 2 mL adapter again, install it into
   the tissue grinder and grind it at 55 Hz for 2 min. 4). Repeat step 3
   twice. 5). Take out the centrifuge tube, centrifuge for 10 min at
   12,000 rpm and 4 °C, take all the supernatant, transfer it to a new
   2 mL centrifuge tube, concentrate and dry it. 6). Accurately add 300 μL
   of 2-Amino-3-(2-chloro-phenyl)-propionic acid (4 ppm) solution prepared
   with 50% methanol water (stored at 4 °C) to redissolve the sample,
   filter the supernatant by 0.22 μm membrane and transfer into the
   detection bottle for LC-MS detection. Liquid chromatography condition:
   The LC analysis was performed on a Vanquish UHPLC System (Thermo Fisher
   Scientific, USA). Chromatography was carried out with an ACQUITY UPLC ®
   HSS T3 (150 × 2.1 mm, 1.8 μm) (Waters, Milford, MA, USA). The column
   maintained at 40 °C. The flow rate and injection volume were set at
   0.25 mL/min and 2 μL, respectively. For LC-ESI (+)-MS analysis, the
   mobile phases consisted of (B2) 0.1% formic acid in acetonitrile (v/v)
   and (A2) 0.1% formic acid in water (v/v). Separation was conducted
   under the following gradient: 0–1 min, 2% B2. 1–9 min, 2% ∼ 50% B2.
   9–12 min, 50% ∼ 98% B2. 12–13.5 min, 98% B2. 13.5–14 min, 98% ∼ 2% B2.
   14–20 min, 2% B2. For LC-ESI (−)-MS analysis, the analytes were carried
   out with (B3) acetonitrile and (A3) ammonium formate (5 mM). Separation
   was conducted under the following gradient: 0–1 min, 2% B3. 1–9 min,
   2% ∼ 50% B3. 9–12 min, 50% ∼ 98% B3. 12–13.5 min, 98% B3. 13.5–14 min,
   98% ∼ 2% B3.14–17 min, 2% B3. Mass spectrum conditions: Mass
   spectrometric detection of metabolites was performed on Q Exactive
   (Thermo Fisher Scientific, USA) with ESI ion source. Simultaneous MS1
   and MS/MS (Full MS-ddMS2 mode, data-dependent MS/MS) acquisition was
   used. The parameters were as follows: sheath gas pressure, 30 arbs. Aux
   gas flow, 10 arbs. Spray voltage, 3.50 kV and − 2.50 kV for ESI (+) and
   ESI (−), respectively. Capillary temperature, 325 °C. MS1 range, m/z
   100–1000. MS1 resolving power, 70,000 FWHM. number of data dependant
   scans per cycle, 10. MS/MS resolving power, 17,500 FWHM. normalized
   collision energy, 30 eV. dynamic exclusion time, automatic.

2.6. Analysis of cellulose content and baijiu yield of fermented grains

   Take 2.5 kg of sorghum, crush the sorghum granules into 4–5 petals by
   crusher, add 75% 90 °C hot water with sorghum quality, mix well and
   soak for 12 h, evenly spread on the upper layer of grate with gauze,
   put it in a boiling water steamer, steam the grain for 40 min,
   immediately take out the grain, add 20% warm water of sorghum quality,
   and add 0.5 kg of rice husk steamed for 30 min to mix well, spread and
   cool. When the grain temperature was lower than 25 °C, 0.75 kg of Daqu
   powder was added to the control group (CT Daqu) and the experimental
   group (FF Daqu). After mixing, it is packed in a ceramic jar and
   subjected to solid fermentation at room temperature. The cellulose
   content was determined by fermentation for 35 days, and the
   colorimetric method was used: cellulose was hydrolyzed to β-glucose by
   heating under acidic conditions. Then, under the action of concentrated
   sulfuric acid, the monosaccharides are dehydrated to form furfural
   compounds. Colorimetric assays can be performed by reacting anthrone
   reagents with the blue-green color of furfural compounds ([64]Xu et
   al., 2023). The linear equation of cellulose is y = 0.0023× - 0.0164,
   R^2 = 0.977.

   After 35 days of solid-state fermentation, the fermented grains were
   distilled (the steamed rice husk was added according to the moisture
   content of fermented grains), and the baijiu head (rich in higher
   alcohols) was removed after 20 mL. The yield of base baijiu (kg) = the
   conversion coefficient of base baijiu sample alcohol content converted
   to 65%vol base baijiu × the quality of base baijiu before conversion.
   Raw material baijiu yield (%) = converted to 65%vol base baijiu yield /
   raw material consumption × 100.

2.7. Data processing

   Use Excel 2019 for raw data collation (Microsoft Corp., Redmond, WA,
   USA), Origin 2023 for histogram graphing (OrginLab., Northampton, MA,
   USA), SAS 9.2 for data statistics and analysis (SAS Inc., Cary, NC,
   USA). Multivariate statistical analysis and mapping were performed
   using R language vegan package, ggplot2 package, ropls package, etc.

3. Results

3.1. Sensory evaluation and physical and chemical property analysis of Daqu

   Sensory evaluation showed that the surface of FF Daqu was yellowish
   brown, with a few cracks, neat cross-section, abundant hyphae, a little
   flocculent color, and pure aroma. The appearance of CT Daqu was
   yellowish brown with a few cracks. The section has rich hyphae, single
   color and no variegated color, pure aroma, both Daqu reached the
   excellent level (Fig. S1C). The physical and chemical properties of
   Daqu are direct indicators of whether Daqu fermentation is mature and
   the quality of Daqu ([65]Deng et al., 2020). As shown in Fig. S2, the
   cellulase activity, acidity, amino acid nitrogen content, saccharifying
   power and liquefaction power of FF Daqu were significantly higher than
   those of CT Daqu. The cellulase activity, saccharification power and
   liquefaction power in the enzyme activity increased by 8.97%, 122.67%
   and 157.57%, respectively. The acidity and amino acid nitrogen content
   in the physical and chemical factors increased by 18.18% and 95.52%,
   respectively. There was no significant difference in moisture content,
   starch content, fermentation ability and esterification ability between
   two kinds of Daqu.

3.2. HS-SPME-GC–MS analysis of Daqu volatile flavor compounds

3.2.1. Composition and difference analysis of Daqu flavor substances

   A total of 60 volatile flavor compounds were identified in these two
   kinds of Daqu, including 16 acids, 13 esters, 10 aldehydes, 7 alcohols,
   4 ketones, 4 phenols, 3 aromatics and 3 others. Among them, 44 and 50
   flavor compounds were identified in FF and CT Daqu, respectively (Table
   S1). The relative content of acid compounds in FF Daqu was
   significantly higher than that in CT Daqu. In addition, the ketone
   compounds in FF Daqu were significantly lower than those in CT Daqu,
   while there was no significant difference in alcohols, esters, phenols,
   aldehydes and aromatic compounds ([66]Fig. 1A). PCA analysis showed
   that the confidence ellipse was not separated within the 95% confidence
   interval, and the flavor composition of the two Daqu was similar (the
   overall difference explained rate was 66.4%). The bioturbation of
   high-yield cellulase strains to strengthen Daqu had little effect on
   the flavor composition of Daqu (PERMANOVA: R^2 = 0.39, p = 0.1)
   ([67]Fig. 1B). In order to further analyze the difference in the
   contribution of flavor substances between the two Daqu samples, the
   variable importance projection (VIP) of OPLS-DA was used to determine
   the candidate differential flavor compounds in Daqu samples. It can be
   seen that Q^2 was 0.4, R^2Y was 0.7, and R^2X was 0.6, indicating that
   the model fitting accuracy was good ([68]Fig. 1C). With VIP value >1 as
   the threshold, as shown in [69]Fig. 1D, 29 flavor substances are the
   key substances for the difference in Daqu flavor.

Fig. 1.

   [70]Fig. 1
   [71]Open in a new tab

   CT and FF Daqu volatile flavor composition classification histogram
   (A), principal component analysis (B), OPLS-DA score diagram (C) and
   volatile flavor substance VIP value diagram (D).

3.2.2. Analysis of flavor contribution compounds of Daqu

   The relative odor activity value (ROAV) was calculated according to the
   relative abundance of the odor threshold binding assay of the compound,
   and the main flavor substance (ROAV ≥1) common to CT and FF Daqu was
   3-Methylbutanoic acid, Hexanoic acid, Phenylethyl alcohol, Hexanoic
   acid ethyl ester and Acetophenone, there was no significant difference
   in the relative content of the above main flavor substances (p > 0.05)
   (Table S1). In addition, the unique main flavor substance of CT Daqu is
   1-Octen-3-ol, Octanoic acid ethyl ester, and the unique main flavor
   substance of FF Daqu is Nonanal (Table S1). Octanoic acid,
   9-Octadecenoic acid ethyl ester, Benzaldehyde are modified compounds
   for the overall flavor of CT Daqu (0.1 < ROAV <1). 2-Methylpropanoic
   acid, Octanoic acid, Benzaldehyde, and Vanillin are modified compounds
   for the overall flavor of FF Daqu (Table S1). The composition and
   content of the main body and modified flavor compounds of the two Daqu
   were not significantly different (p > 0.05).

3.3. Analysis of the composition and differences of active microbiota of the
Daqu

   Metatranscriptomics was utilized to explore the active microorganisms
   of mature Daqu, and 43.53 Gbp of raw data were obtained from
   288,295,034 reads, in which the percentage of ambiguous bases was low
   (0.002%), the GC accounted for about 45% of the total, and the
   percentage of bases with base identification accuracy of 99% or more
   (Q20) were higher than 92%, and both the amount of sequencing and the
   high quality base percentage indicated that the overall quality of
   sequencing was high (Table S2). Sequences that were too short in
   length, contained too many ambiguous bases and adulterated with
   adapters (Adapter) were further screened and filtered to obtain 88.34%
   of valid reads yielding 38.17 Gbp of clean data for subsequent analysis
   (Table S2). In addition, splicing was performed with MEGAHIT, 559630
   non-redundant contigs with an N50 length of 642 bp were obtained, which
   were then predicted to 10,089 non-redundant proteins (Table S3).

   Using Kaiju annotated to 151 phylums, 233 orders, 607 orders, 1350
   families, 4124 genera, and 15,042 species. The relative abundance of
   dominant phyla (relative abundance >1%), dominant genera, and dominant
   species of the FF Daqu and CT Daqu accounted for 98.43%, 97.34%,
   95.79%, 91.50%, and 85.08%, 63.42%, respectively, of all active
   microorganisms. As shown in [72]Fig. 2A, fungi predominate in the
   active microbial community. The dominant phylum of FF Daqu is
   Ascomycota and Pisuviricota ([73]Fig. 2B), the dominant genus is
   Aspergillus, Paecilomyces, Penicillium ([74]Fig. 2C), and the dominant
   species are Paecilomyces variotii, Aspergillus chevalieri, Aspergillus
   cristatus, Aspergillus glaucus, Aspergillus ruber, Aspergillus wentii,
   Penicillium steckii, Penicillium decumbens ([75]Fig. 2D). The dominant
   phylum of CT Daqu is Ascomycota ([76]Fig. 2B), and the dominant genus
   is Aspergillus, Paecilomyces, Penicillium, Talaromyces and Rasamsonia
   ([77]Fig. 2C), the dominant species are Paecilomyces variotii,
   Aspergillus chevalieri, Aspergillus cristatus, Aspergillus glaucus,
   Aspergillus ruber, Aspergillus wentii, Penicillium steckii, Penicillium
   decumbens, Rasamsonia emersonii, Aspergillus terreus, Ophiocordyceps
   lanpingensis, Aspergillus sclerotialis, Penicillium brasilianum,
   Aspergillus sp. HF37, Aspergillus pseudoglaucus, Metarhizium majus
   partitivirus 1 ([78]Fig. 2D). As shown in [79]Fig. 2E, the principal
   coordinate analysis explained 99.8% of the differences in the original
   data on the PCoA1 axis, indicating that the species composition spectra
   of FF and CT Daqu were different, and UPGMA hierarchical clustering
   analysis expressed that the two Daqu sample groups had a high degree of
   similarity. By analyzing the difference of ANOSIM (Analysis of
   similarities) at the species level of the two Daqu samples, the R value
   was 1, indicating that the difference between the groups was large and
   the difference within the group was small. In order to test whether the
   difference was statistically significant, 999 substitution tests were
   performed based on PERMANOVA, and the P value was 0.1 > 0.05, and the
   difference between FF and CT Daqu groups was not significant,
   indicating that functional strain strengthening did not significantly
   change the structure of Daqu active microbiota.

Fig. 2.

   [80]Fig. 2
   [81]Open in a new tab

   Composition of active microbial communities of CT and FF Daqu (A:
   boundary level, B: phylum level, C: genus level, D: species level),
   species composition PCA and HCA analysis (E), microbiota composition
   difference analysis (F, G) and LEfSe analysis (H).

   By normalizing the composition abundance of each classification level
   of each sample, the abundance difference of each classification unit of
   active microorganisms between sample groups was carried out. As shown
   in [82]Fig. 2F and G, the classification of FF and CT Daqu samples was
   highly consistent at the phylum level, and the different species were
   mainly classified into Eurotiales, Mycosphaerellales, Pleosporales,
   Helotiales, Saccharomycetales, Hypocreales, and Xylariales. Compared
   with CT Daqu, FF Daqu was significantly upregulated in Bacillales,
   Enterobacterales, Boletales, and Chaetothyriales, Pezizales, and
   Diaporthales species were significantly down-regulated. The
   nonparametric Kruskal-Wallis and Wilcoxon rank sum tests were used to
   perform simultaneous differential analysis for all classification
   levels combined with Linear discriminant analysis (LDA) effect size. h3
   (Paecilomyces variotii) and e2 (Aspergillus cristatus) are the
   taxonomic units with the largest horizontal nodes of the FF and CT Daqu
   taxonomic clades, respectively, and are biomarkers reflecting the
   differences between the two large curves ([83]Fig. 2H).

3.4. Carbohydrate-active enzyme (CAZy) composition and its contributing
microbial analysis

   The search module using MMseqs2 was used to compare and annotate 10,089
   non-redundant proteins with commonly used protein databases. The
   abundance of each protein corresponding to the CAZy enzyme family was
   statistically annotated to different modules. The results showed that
   the abundance of glycoside hydrolase (GH) and oxidoreductase (AA) was
   significantly higher in FF Daqu than CT Daqu, and the abundance of
   polysaccharide cleavage enzyme (PL) and carbohydrate-binding module
   (CBM) was significantly lower than that of CT Daqu.
   glycosyltransferases (GTs) and glycoesterases (CEs) did not differ
   significantly between the two types of Daqu ([84]Fig. 3A). A total of 9
   enzymes were annotated in the PL family, and PL3_2 and PL1_1 were the
   most abundant enzymes in FF Daqu and CT Daqu, respectively ([85]Fig.
   3B). A total of 11 enzymes were annotated in the CE family, and CE5 and
   CE8 were the enzymes with the largest proportion of FF Daqu and CT
   Daqu, respectively ([86]Fig. 3C). A total of 85 enzymes were annotated
   in the GH family, and the enzymes with the highest abundance in FF Daqu
   and CT Daqu were GH19 and GH18, respectively ([87]Fig. 3D). A total of
   51 enzymes were annotated in the GT family, and the most abundant
   enzyme in the two Daqu was GT4 ([88]Fig. 3E). There were 26 annotations
   in the CBM family, and CBM13 accounted for the highest proportion of
   the two large curved species ([89]Fig. 3F). There were 16 enzymes in
   the AA family, AA1 accounted for the highest proportion in FF Daqu and
   AA10 accounted for the highest proportion in CT Daqu ([90]Fig. 3G).

Fig. 3.

   [91]Fig. 3
   [92]Open in a new tab

   Carbohydrate active enzyme (CAZy) family abundance comparison (A) and
   composition (B: PL, C: CE, D: GH, E: GT, F: CBM, G: AA family) and some
   enzyme comparison (H) analysis.

   Among them, the GH enzyme family is classified by function as amylase
   (GH13, GH13_1, GH15, GH57), cellulase (GH1, GH2, GH3, GH5, GH6, GH7,
   GH9, GH12, GH16, GH30, GH51, GH131), and hemicellulase (GH5). And the
   CE enzyme family includes carboxylesterase (CE4, CE12), pectin methyl
   esterase (CE8). Further comparing the abundance of the above enzymes in
   the two Daqu samples, the abundance of amylase, cellulase and
   hemicellulase in FF Daqu was significantly higher than that in CT Daqu,
   and carboxylesterase and pectin methyl esterase were lower than that in
   CT Daqu ([93]Fig. 3H). These results were more in line with the results
   of the analysis of the Daqu's physicochemical properties.

   Amylase producers differed between FF and CT Daqu, with the main
   sources in FF Daqu being Aspergillus chevalieri, Trichomonascus
   ciferrii, Byssochlamys spectabilis, Penicillium decumbens, Penicillium
   steckii, and in the CT Daqu mainly from Aspergillus cristatus,
   Aspergillus ruber, Macrophomina phaseolina (Table S4). Cellulase,
   carboxylesterase and pectin methyl esterase contributing microorganisms
   were more widespread, and overall, the source of the above enzymes in
   FF Daqu was mainly Penicillium, while CT Daqu was mainly biased towards
   Aspergillus (Table S4). In addition, hemicellulase was only derived
   from Byssochlamys spectabilis in FF Daqu (Table S4).

3.5. Daqu non-volatile metabolite metabolomic analysis

   As shown in [94]Figs. 4A and B, PCA based on reliable metabolites
   explained 48.8% and 46.2% of the total variation in ESI+ and ESI- mode,
   respectively. FF Daqu and CT Daqu had obvious separation in both modes,
   and the cumulative explanatory rates (R^2X) of the PC1 axis direction
   model were 0.569 and 0.551, respectively, indicating that the goodness
   of fit of the model was good. The within-group error is further ignored
   and the random error irrelevant to the research purpose is eliminated,
   and a supervised orthogonal-partial least squares discriminant analysis
   (OPLS-DA) model is constructed, so as to maximize the intergroup
   differences. The OPLS-DA model had good reliability and predictability
   in the ESI+ model: R^2Y = 1, Q^2 = 0.976 and in the ESI- model:
   R^2Y = 1, Q^2 = 0.939, and the results showed significant differences
   between the two Daqu metabolites (PERMANOVA: R^2 = 0.89, p = 0.014). At
   the same time, the displacement test shows that the OPLS-DA model does
   not have overfitting and can be used for subsequent screening of
   characteristic metabolites between different taxa.

Fig. 4.

   [95]Fig. 4
   [96]Open in a new tab

   PCA, OPLS-DA and their model permutation test analyses of metabolite
   composition of FF Daqu and CT Daqu (A, B), hierarchical cluster
   analysis (C), volcano plot of differential metabolites (D) and
   metabolic pathway enrichment bubble plot (E).

   Hierarchical cluster analysis showed that 12 classes of metabolites
   were classified according to chemical structure and secondary
   classification, which dominated the difference between two kinds of
   Daqu. In FF Daqu, Alcohols and polyols, Amino acids, peptides, and
   analogues, Purines and purine derivatives, Lineolic acids and
   derivatives, Indoles, Fatty acids and conjugates were higher than CT
   Daqu, Pyridinecarboxylic acids and derivatives, Methoxyphenols,
   Lineolic acids and derivatives, Indoles, Carbonyl compounds, Benzoic
   acids and derivatives, Amines, Carbohydrates and carbohydrate
   conjugates were lower than CT Daqu ([97]Fig. 4C). Projected with the
   importance value of the first principal component variable of OPLS-DA:
   VIP >1 and statistical significance (p < 0.05), FF Daqu and CT Daqu
   obtained 250 and 108 differential metabolites from ESI+ and ESI- modes,
   respectively, and the differential volcano plot showed that FF compared
   with CT upregulated metabolites 184 and downregulated metabolites 174,
   and the first 5 metabolites with the smallest P value were upregulated
   as 1,7-Dimethyluric acid, L-threo-3-Phenylserine,
   N-Acetyl-L-phenylalanine, downgraded to Miglitol, Glucose 6-phosphate
   ([98]Fig. 4D).

   KEGG PATHWAY enrichment analysis was performed on KEGG (Kyoto
   Encyclopedia of Genes and Genomes, [99]http://www.genome.jp/kegg/) by
   differential metabolite comparison, and a total of 209 pathways were
   enriched ([100]Fig. 4E).

3.6. Correlation analysis of dominant active microbiota with physicochemical
properties, flavor contributing compounds (ROAV>0.1) and differential
metabolites

   The difference in physicochemical properties leads to changes in the
   microbiota, which will change the enzymatic properties of Daqu. For the
   entire fermentation system, changes in enzymatic properties lead to
   differences in metabolic levels, which in turn affect flavor and change
   physicochemical properties. Correlations between 14 active
   microorganisms and physicochemical properties, 23 flavor-contributing
   compounds and 50 differential metabolites were analyzed using
   Spearman's correlation test. For physicochemical properties, the
   correlation network showed that the active microorganisms were
   significantly correlated with Amino nitrogen, Cellulose activity,
   Acidity, Liquefying power, and Saccharifying power (r^2 > 0.5,
   p < 0.05). Of these, there were 10 degrees between Acidity and the
   active microorganisms, of which significant positive correlations were
   found between Metarhizium majus partitivirus 1, Penicillium steckii,
   Penicillium decumbens, Paecilomyces variotii. Saccharifying power and
   Liquefying power showed significant positive correlation with
   Penicillium steckii and Penicillium decumbens, respectively, while
   other physicochemical properties showed significant negative
   correlation with active microorganisms ([101]Fig. 5A). For
   flavor-contributing compounds, only 9 compounds were significantly
   correlated with active microorganisms. Hexadecanoic acid, ethyl ester
   is the largest node in the flavor network, and is significantly
   positively correlated with 8 active microorganisms such as Metarhizium
   majus partitivirus 1, Aspergillus chevalieri, Aspergillus terreus, and
   Aspergillus sclerotialis. Acetophenone, 9-Octadecenoic acid ethyl
   ester, Benzyl alcohol, Phenylethyl alcohol, Benzaldehyde and microbiota
   were significantly positively correlated, while Acetic acid and Phenol
   were mainly negatively correlated ([102]Fig. 5A).

Fig. 5.

   [103]Fig. 5
   [104]Open in a new tab

   Correlation analysis of species-level active microorganisms with
   physicochemical properties, flavor substances (A), and nonvolatile
   metabolites (B). AN: Amino nitrogen, CA: Cellulose activity, LP:
   Liquefying power, SP: Saccharifying power.

   For non-volatile metabolites, L-threo-3-Phenylserine, Pyridoxal
   5′-phosphate, Aspartame, D-Octopine, L-Arginine,
   N-Acetyl-L-phenylalanine, 5’-Deoxyadenosine, Glucose 6-phosphate,
   Uridine, and Styrene Oxide showed significant correlation with the
   active microorganisms. Metarhizium majus partitivirus 1, Aspergillus
   terreus and Aspergillus sclerotialis had a high degree of connection
   with metabolites, accounting for 32.57% of the total.
   3-Methyl-2-oxovaleric acid, 4-Methylbenzaldehyde, Capsaicin,
   Digalacturonate, Fructoselysine 6-phosphate, gamma- Aminobutyric acid,
   and Tryptophanol were significantly negatively correlated with active
   microorganisms only ([105]Fig. 5B).

3.7. Comprehensive analysis of metabolites and enzymes

   In order to reconstruct the active microbial metabolic network of
   mature Daqu, a metabolic map covering non-volatile metabolites and
   active microbial enzymes was constructed based on the public database
   (KEGG) and the results of pathway analysis and enrichment analysis of
   differential metabolites ([106]Fig. 6). Based on the actual production
   of Daqu and the needs of brewing, the conversion of carbohydrates and
   amino acids was comprehensively analyzed. In Starch and sucrose
   metabolism, amylose generates starch under the action of 1,4-α-glucan
   branching enzyme [EC: 2.4.1.18], and then forms dextrin under the
   action of α-amylase [EC: 3.2.1.1], and finally generates d-Glucose
   under enzymatic action. The source of d-Glucose was also hydrolyzed by
   beta-fructofuranosidase [EC: 3.2.1.26] to hydrolyze sucrose and
   cellulose. Endoglucanase [EC: 3.2.1.4], cellulose 1,4-β-cellobiase [EC:
   3.2.1.91] and β-glucosidase [EC: 3.2.1.21] were up-regulated in FF
   Daqu, and the content of cellobiose was lower than that in CT Daqu,
   indicating that the ability of cellulose hydrolysis in fortified Daqu
   was higher. Various intermediates of the glycolysis process are
   crosslinked with other metabolic pathways, such as: β-d-Fructose-6P,
   Glycerate-3P, Phosphoenolpyruvate. The first key enzyme of glycolysis,
   hexokinase [EC: 2.7.1.1], was highly expressed in fortified Daqu,
   catalyzing glucose to form α-d-Glucose-6P, and then formed more
   β-d-Fructose-6P under the action of highly expressed phosphohexose
   isomerase [EC: 2.7.1.1]. In pectin metabolism, the degradation of
   D-galacturonic acid mediated by fungi in CT Daqu was higher, and
   endopolygalacturonase [EC: 3.2.1.15] catalyzed the formation of
   Digalacturonate. Digalacturonate is hydrolyzed to D-Galacturonate by
   galactouronic acid 1,4-α-galacturonase [EC: 3.2.1.67], and the methoxy
   group is removed to form methanol. In terms of alcohol metabolism, the
   expression levels of alcohol dehydrogenase EutG, [EC: 1.1.1.2] and [EC:
   1.1.2.8] in FF Daqu were higher than those in CT Daqu. In fatty acid
   metabolism, the expression of long-chain acyl-CoA synthetase [EC:
   6.2.1.3] in CT Daqu was higher than that in FF Daqu, which catalyzed
   Hexadecanoate to form Hexadecanoyl-CoA. Therefore, the content of
   Hexadecanoate in FF Daqu was higher than that in CT Daqu.

Fig. 6.

   [107]Fig. 6
   [108]Open in a new tab

   Comprehensive metabolic pathways of non-volatile metabolites and
   enzymes. The left side of the small rectangle of the black border
   represents the content of the metabolite in FF Daqu, and the right side
   represents the content of the metabolite in CT Daqu. Red represents the
   higher content of the substance in the Daqu. The yellow rectangle and
   the blue rectangle represent the up-regulated expression and
   down-regulated expression of the enzyme, respectively. (For
   interpretation of the references to color in this figure legend, the