Introduction

For millennia, humans have continuously optimized the growth and metabolic conditions of specific microbial communities. This is best exemplified for food fermentations, an ancient food production method which through repeated production practices resulted in the repetitive enrichment and eventual domestication of certain microorganisms responsible for defining flavor and quality characteristics1,2,3. However, the core microorganisms required for producing many traditional fermented foods and their specific contributions to the final food products remain to be determined.

Huangjiu is one of China’s most popular alcoholic beverages and archeological records suggest this beverage has been made for nearly nine millennia4,5. Huangjiu is made as a result of a complex, spontaneous, mixed-culture process in an open-air environment. There are two main methods for huangjiu fermentation, traditional method and industrial method. The principles of these two methods are similar, except for producing time and some materials. In traditional method, wheat qu is prepared in summer, materials are prepared in winter, and the overall fermentation needs almost one year, while in industry materials can be prepared in anytime and the fermentation is much faster. Meanwhile, comparing to traditional fermentation, industrial fermentation needs additional Aspergillus flavus SU16. The industrial huangjiu production process usually involves several processes wherein the rice is soaked, steamed, and a two-stage huangjiu fermentation process (Fig. 1). Saccharomyces cerevisiae, wheat qu, steamed glutinous rice, and water were the major ingredients for huangjiu fermentation, and the mixture is called huangjiu fermentation mash (HJFM). The wheat qu starter culture which is made by wheat fermentation affects the flavor quality of the final product and contains Aspergillus, Rhizopus, and other fungi6,7. The primary contribution of S. cerevisiae is ethanol production8, while the primary contribution of wheat qu is to provide liquefaction and saccharifying power, which also affects the flavor quality of the final product. Although wheat qu is thus a key ingredient in ancient alcoholic huangjiu fermentation, the core microbiota of them still remains indistinct. Additionally, Aspergillus flavus, a filamentous fungus, has been widely used in large-scale production of traditional fermented food products, such as sake, miso, soy sauce, and huangjiu9,10. A. flavus SU-16 (also known as Aspergillus oryzae SU16) is often inoculated into cooked wheat qu and used as a supplement of enzyme preparation in huangjiu fermentation9. Although A. flavus SU-16 have been used for industrial wheat qu preparation11, the aroma characteristics of resulting huangjiu are significantly different from traditional huangjiu5. Therefore, there remains the need to identify the core microbiota of wheat qu and huangjiu responsible for the characteristics of the traditional huangjiu beverage.

Fig. 1: Manufacture chart of wheat qu and HJFM.
figure 1

Wheat qu Production Process: (I) Crushing, Mixing and Shaping. The wheat is crushed, blended with water (~20% V/W), and then pressed into molds of different size depending on the manufacture, which are called wheat qu blocks. (II) Fermentation. the wheat qu blocks are moved into qu house for culturing under ventilated and insulated conditions. Fermentation is divided into three fermentation stages: First, 30 to 40 °C for 3 days, which results in the mass reproduction and growth of microorganisms. Second, 45 to 55 °C for 4 to 5 days, which is used mainly for the accumulation of microbial metabolites. Third, 10 to 12 days above 40 °C, which is conducive to moisture removal in the starter core and the formation of aroma substances. The wheat qu were dried to a water content below 12% weight. (III) Storage. After cooling downing to room temperature, the wheat qu was stored for 60 to 90 days. Huangjiu Production Process: (I) Soak rice. The rice is washed and soaked in natural water at room temperature for 40 h. (II) Steamed rice. The soaked rice is steam-cooked in a steaming machine to gelatinize the starch contained in the rice grains. (III) Mixing. The steamed rice is mixed with wheat qu, water and S. cerevisiae at a weight ratio of 15:1.4:12.5:1.1, and the mixture is called huangjiu fermentation mash (HJFM). (IV) This process can be divided into two stages. The stage I fermentation temperature is 28–32 °C and lasts for 3–4 day, and stage II lasts 15 to 20 days and the temperature controlled below 15 °C. (IV) Pressing, sterilization and storage. After fermentation, the HJFM were filtered by pressing. Sterilize and store the fresh huangjiu.

Full size image

Flavor compounds are very important indicators of huangjiu’s quality12,13. Currently, more than 900 kinds of volatile flavor compounds have been detected in huangjiu, including acids, alcohols, aldehydes, esters, ketones, and phenols. The flavor compounds are greatly affected by microorganisms in huangjiu brewing, while the formation mechanism is still poorly understood. Besides its unique flavor compounds, huangjiu is also known to contain biogenic amines (BAs)14. These compounds are low molecular weight, non-volatile nitrogenous organic bases made from the decarboxylation of amino acids15. BAs contribute to the normal physiological function of the human body, however, high intake may also result in the adverse physiological reactions of headache, nausea and other allergic reactions3,12. The rice soaking and huangjiu fermentation process result in increasing concentrations of the BAs, including histamine, tyramine, putrescine and cadaverine16. Although the quantities of BAs in huangjiu (39.30–241 mg/L) are reportedly below levels known to cause adverse reactions14,17, recent studies have shown that BAs are responsible for the unpleasant “hangover” symptoms after drinking huangjiu12. Reducing the BAs concentration is a viable way to improve the quality and taste of huangjiu, and other fermented food products that contain high concentration of BAs, such as sausage and soy sauce.

Accordingly, in this study, a comprehensive systems biology approach was applied using metagenome sequencing, metabolomics, metabolic pathways, and metabolites analysis to identify the core functional microorganisms in wheat qu and huangjiu fermentations. This approach resulted in the discovery of Saccharopolyspora as a key member of the huangjiu microbiome and the role of members of this genus in BAs metabolism. Meanwhile, microbial strains that were involved in degradation of BAs in wheat qu and huangjiu fermentation was isolated, and its potentials and metabolic activities to reduce BAs in huangjiu were evaluated by co-culture fermentation for the first time, as well as other fermentations with high concentration of BAs, such as sausage and soy sauce. The analysis showed us in-depth views of core functional microorganisms in huangjiu fermentations, and provided a viable way to reduce the BAs in huangjiu and other fermented food.

Results

Saccharopolyspora is abundant in the wheat qu and HJFM microbiota

To uncover the variation curve of microorganisms during fermentation, microbial DNA copy numbers and community structure were analyzed using quantitative real-time PCR (qPCR) and metagenomic analyses. DNA copy numbers of total bacteria and fungi in wheat qu were 11.2 ± 0.2 log10 copies/g and 6.6 ± 0.1 log10 copies/g. Metagenome sequencing showed a total of 430 genera (including 263 bacteria and 167 fungi) and 1442 (including 1176 bacteria and 266 fungi) species in the wheat qu sample obtained from a commercial huangjiu factory (Supplementary Table 1). Bacteria were dominant, constituting 90.24% of the microorganisms present and were predominantly in the Actinobacteria (86.66%) phylum (Fig. 2A). Only six genera were detected at >1% abundance, and Saccharopolyspora (72.28%), and Streptomyces (2.29%) were the most represented. Other taxa included Bacillus (1.41%), Pseudomonas (1.56%), Amycolatopsis (1.03%) and Lichtheimia (1.14%). Metagenome sequences were also able to resolve species-level differences. S. hirsuta was the most abundant (38.80%), followed by other species of Saccharopolyspora, including S. rectivirgula (12.71%), S. shandongensis (5.42%), S. erythraea (2.59%), and S. spinosa (2.50%).

Fig. 2: Microbial communities in wheat qu and huangjiu fermentation.
figure 2

A Microbial communities of wheat qu at phylum (inner circles), genus (intermediate circle) and species levels (outer circles). B Dynamics of functional microorganism during huangjiu fermentation process at species level.

Full size image

DNA copy numbers of total bacteria and total fungi in HJFM rapidly increased to 6.70 ± 0.1 log copies/g and 5.81 ± 0.0 log copies/g in the first 25 h of fermentation, and reached 12.55 ± 0.3 and 8.90 ± 0.3 log copies/g at 50 h, respectively. Bacterial DNA copy number was maintained at 12.55 to 11.72 log copies/g between 50 and 100 h of fermentation, while fungal DNA copy number gradually increased from 8.90 to 10.77 log copies/g between 50 and 100 h of fermentation. At 300 h, bacterial and fungal DNA copy numbers decreased to 8.60 ± 0.3 log copies/g and 7.34 ± 0.1 log copies/g, respectively. At the 25, 50, 100 and 300 h time points during the fermentation, the microbial contents of the HJFM were mainly comprised of Actinobacteria, Firmicutes, and Ascomycota phyla (Supplementary Table 2). The relative abundance of the dominant genus, Saccharopolyspora, increased from 40.37% to 51.48% during huangjiu fermentation process (Fig. 2B and Supplementary Table 2). A variety of Saccharopolyspora species were detected including S. rectivirgula, S. erythraea, S. hirsuta, S. antimicrobica, S. spinosa, and S. shandongensis (Fig. 2B and Supplementary Table 2). Saccharomyces, specifically S. cerevisiae, was also present in the HJFM, and reached the highest levels (22.28%) in the HJFM 100 h post-incubation and then declined to 11.34%. Aspergillus was also present in more modest amounts (9.82%) after 25 h incubation, but declined over time to 4.17% of the total microbial contents by 300 h.

Metabolites profiling of wheat qu and HJFM

The abundant metabolites were the most outstanding feature of Huangjiu, comparing to other fermentation fermented alcoholic beverage. In this study, A total of 103 metabolites were identified in the wheat qu sample tested, including 8 organic acids, 18 amino acids, and 77 volatile compounds. Tartrate (2.99 g/kg wheat qu), lactate (2.65 mg/kg wheat qu) and acetate (1.91 mg/kg wheat qu) accounted for 81.53% (9.26 mg/kg wheat qu) of the 8 organic acids detected (Supplementary Fig. 1A). Pro, Glu, Arg, Asp, Ala and Lys were the most abundant amino acids in the wheat qu, comprising 70.14% (744.48 mg/kg wheat qu) of the 18 amino acid quantities detected (Supplementary Fig. 1B). The total content of the volatile compounds in wheat qu were found to be 1762.37 μg/kg wheat qu. Among the identified volatile compounds, alcohols (phenylethanol, 1-octen-3-ol, 2,3-butanediol, 2-octene-1-ol) and esters (hexanoate ethyl, ethyl palmitate, ethyl acetate, dibutyl phthalate) were the most abundant (relative abundances of 65.7% and 14.5%, respectively).

Metabolites present in the HJFM changed over time (Supplementary Fig. 1). Lactate, tartrate, and citrate were the dominant organic acids in HJFM, comprising between 86.91% to 90.00% of the 8 organic acids identified (Supplementary Fig. 1A). Seven amino acids (Arg, Try, Ala, Pro, Glu, Leu and Gly) accounted for between 58.42% (400 h) to 70.19% (100 h) of the 18 amino acids detected (Supplementary Fig. 1B). 64 volatile compounds were also identified in the HJFM, 5 alcohols (phenylethanol, isoamylol, isobutanol, 3-methylthiopropanol, and 2,3-butanediol), 10 esters (ethyl lactate, ethyl acetate, diethyl succinate, ethyl butyrate, isoamyl acetate, isobutyl acetate, phenethyl acetate, ethyl decanoate, ethyl octanoate, and ethyl palmitate) and 3 phenols (4-vinylguaiacol, 4-ethyl guaiacol, and 4-ethyl phenol) were the main volatile compounds of HJFM, comprised 89.28% to 99.49% of the total content (Supplementary Fig. 1C). Among the identified volatile compounds, the aroma compounds phenylethanol, isoamylol, isobutanol, 3-methylthiopropanol, ethyl lactate, ethyl acetate, 4-vinylguaiacol, and 4-Ethyl guaiacol were the most abundant (relative abundances from 87.16% to 97.47%) in HJFM.

Identification of core functional microorganisms in HJFM microbiota by metabolic pathways analysis

To identity the core functional microorganisms, metabolic pathway analysis was performed. The HJFM metagenomes confirmed that Saccharopolyspora was likely the core functional microorganism in huangjiu fermentation (Supplementary Fig. 2). Pathway analysis (Supplementary Fig. 2A-a) showed that the enzymes (EC 3.2.1.4 and 3.2.1.91) responsible for cellulose metabolism were highly enriched in Saccharopolyspora (43.09%) and Amycolatopsis (42.78%). Aspergillus (14.03%) was also associated with that function in the metagenomes. Many microorganisms, including Saccharopolyspora (71.05%) and Saccharomyces (19.20%) had the capacity to participate in extracellular starch degradation (Supplementary Fig. 2B-a). The grain proteins of raw material can be hydrolyzed into amino acids or peptides by the action of enzymes (EC 3.4.--, 3.4.11.1, 3.4.11.2, 3.4.13.-, 3.4.16.4, and 3.4.17.14) in HJFM metagenomes, and 10 microorganisms may be involved in protein degradation, among which, Saccharopolyspora (82.88%) might be the main actors.

Phenylethanol, isoamylol and isobutanol are the most abundant higher alcohols in wheat qu and HJFM. In both catabolic (Ehrlich pathway) and biosynthetic pathways, α-ketoic acids are converted into higher alcohols by carboxylases (EC 4.1.1.-) and dehydrogenases (EC 1.1.1.1, 1.1.1.2, and 1.1.1.90) (Supplementary Fig. 2A) in HJFM microbiota. Saccharomyces (36.48%), Saccharopolyspora (33.33%), Actinopolyspora (8.63%), and Amycolatopsis (5.87%) may also be involved in higher alcohol synthesis (Supplementary Fig. 2B-c).

Ethyl esters and acetic esters, which are mostly fruity and floral, are important esters in HJFM. Lipases (EC 3.1.1.1, EC 3.1.1.3), the key ester biosynthetic enzymes during fermentation, can catalyze the reversible condensation of acid and alcohols in this study (Supplementary Fig. 2A). As shown in Supplementary Fig. 2B-b, many microorganisms, including Saccharomyces (59.75%), Saccharopolyspora (25.23%) and Staphylococcus (10.66%) had the capacity to synthesize esters. Among them, Saccharomyces and Saccharopolyspora might provide the greatest ester biosynthetic capacity.

Pro, Glu, Arg and Ala were the main amino acids in wheat qu and HJFM, and a number of genes that may participate in amino acid metabolism were observed in the study. Ornithine cyclodeaminase (EC 4.3.1.12) can transform ornithine to Pro, while pyrroline-5-carboxylate reductase (EC 1.5.1.2) can catalytic synthesis of Pro with L-glutamate-5-semialdehyde as the precursors (Supplementary Fig. 2A). Arg can be synthesized from citrulline or l-argininosuccinate by enzymes (EC 3.5.3.6, 1.14.13.39, 1.14.14.47, and 4.3.2.1). Transaminases (EC 2.6.1.1, EC 2.6.1.2) and dehydrogenases (EC 1.4.1.2, 1.4.1.3, and 1.4.1.4) can catalyze the formation of Glu from 2-oxoglutarate. Glu can also be synthesized from l-glutamate-5-semialdehyde using L-glutamate gamma-semialdehyde dehydrogenase (EC 1.2.1.88). Ala is synthesized from pyruvate in a reaction catalyzed by transaminases (EC 2.6.1.2, 2.6.1.21) and alanine dehydrogenase (EC 1.4.1.1). Besides, Asp can be converted into Ala by aspartate 4-decarboxylase (EC 4.1.1.12). Saccharopolyspora (69.92%) and Saccharomyces (20.20%) may also be involved in amino acids synthesis (Supplementary Fig. 2B-d).

Lactate, tartrate, acetate and citrate were the main organic acids in wheat qu and HJFM. Based on the result of metagenome annotation, pyruvate (related enzymes EC 1.1.99.40, 1.1.1.27, 1.1.1.28, 1.1.5.12) is the immediate precursors of lactate synthesis (Supplementary Fig. 2A). It was determined that those related enzymes genes were encoded by Saccharopolyspora (58.90%), Saccharomyces (18.23%), and Staphylococcus (5.06%) (Supplementary Fig. 2B-e). Saccharopolyspora (80.37%) have the capacity to transform oxaloacetate into tartrate using tartrate dehydrogenase/decarboxylases (ttuc, EC:1.1.1.93, 4.1.1.73, and 1.1.1.83). The acetate formation pathways vary across different microbial genera, and Saccharopolyspora (76.63%) and Saccharomyces (13.66%) are the main acetate producers. Citrate is a well-known intermediate in the TCA cycle. Saccharopolyspora (54.13%) and Saccharomyces (38.42%) may participate in the formation of citric acid in HJFM.

In summary, Saccharopolyspora should be a major player in raw material (cellulose, starch and protein) degradation and the synthesis of all the main metabolites. A total of 122 Saccharopolyspora genes encoding flavor-forming enzyme were identified, and most of the genes were present at high abundance. Saccharomyces may participate in starch degradation, and the synthesis of esters, higher alcohols, amino acids, lactate, acetate and citrate. Aspergillus may be involved in cellulose degradation.

O2PLS-based statistical correlations between core functional microorganisms and flavor metabolites

To explore the correlation between the microbiota and main metabolites, a bidirectional orthogonal partial least squares (O2PLS) method was used to analyze the change in the relative abundance of core functional microorganisms (only those with relative abundance ≥1%) and main flavor metabolites (total 7 amino acids, 3 organic acids and 18 volatile compounds) during the fermentation of HJFM over time to explore the correlation between the microbiota and main flavor metabolites (Supplementary Table 3). According to the model, the proportions of five microbial species (VIP(pred) > 1.0) (including four bacterial (S. rectivirgula, S. erythraea, S. hirsuta and S. antimicrobica) and one fungal (S. cerevisiae) species) had a significant influence on the flavor metabolites (Fig. 3A). Among those taxa, the O2PLS model showed that Saccharopolyspora was the most important contributor to the production of flavor metabolites during huangjiu fermentation. Based on the correlation coefficient between microbiota and flavor metabolites (Fig. 3B), S. rectivirgula, S. erythraea, S. hirsuta, S. antimicrobica and S. cerevisiae was positively correlated with 6, 16, 14, 2 and 5 of the main flavor metabolites, respectively (ρ > 0.7). Among them, phenethyl alcohol, isobutyl acetate, ethyl octanoate and phenethyl acetate were only associated with S. cerevisiae proportions (ρ > 0.7). This result indicated that both S. erythraea and S. hirsuta shared a positive correlation with five amino acids (Ala, Arg, Leu, Try, and Glu), two organic acids (tartrate and citrate), and six volatiles (ethyl acetate, diethyl succinate, ethyl butyrate, ethyl lactate, 4-vinylguaiacol, and 4-ethyl guaiacol). The proportions of S. rectivirgula were positively correlated with six flavor metabolites (ρ > 0.7, Pro, Gly, Leu, ethyl lactate, isoamyl acetate, and 4-vinylguaiacol), while S. antimicrobica was only positively correlated with diethyl succinate and 4-ethyl guaiacol. In addition, the proportions of Lactiplantibacillus plantarum, Aspergillus oryzae, S. gallinarum were strongly negatively correlated with 7, 13, and 13 main flavor metabolites, respectively, indicating a potential role in flavor metabolite degradation. In conclusion, based on this model with this single HJFM measured over time, S. erythraea, S. hirsuta, S. rectivirgula and S. cerevisiae may be the core metabolites producing microorganisms in huangjiu fermentations.

Fig. 3: Correlation analyses between core functional microorganisms and main metabolites by O2PLS modeling during the HJFM process.
figure 3

A VIP(pred) (variable importance for predictive components) plot of the core functional microorganisms (only those with relative abundance ≥1%) (B) Pheatmap of the main metabolites and the core functional microorganisms during the HJFM process. The correlation coefficients are represented by the color. Red indicates positive correlation and blue indicates negative correlation. Correlations were calculated based on relative abundance and absolute quantification. Clustering analysis was performed using the Pearson correlation coefficient and Euclidean distance.

Full size image

Enzymatic activities of Saccharopolyspora isolated from wheat qu

To better understand the role of Saccharopolyspora in huangjiu fermentations, we isolated bacteria from the wheat qu. Isolated strains were identified at the species level by 16 S rRNA gene sequencing and found to be S. hirsuta (J1 to J3), S. rectivirgula (J4 and J5), S. shandongensis (J6 to J8) and S. erythraea (J9). Among the isolates, S. hirsuta J2 showed the highest activity of amylase, glucoamylase and protease (Supplementary Fig. 3). The protease and glucoamylase activities of S. hirsuta J2 were 2927.74 ± 16.13 U/g and 2.41 ± 0.06 U/g, respectively, which were significantly higher than those in the control group (P < 0.05), but there were no significant differences in amylase activity (P < 0.05). The amylase, glucoamylase and protease activities of A. flavus SU-16 were significantly higher than control group (P < 0.05). The glucoamylase and protease activities of A. oryzae MQ were significantly higher than control group (P < 0.05), while there was no significant difference in protease activity. And the three enzyme activities of L. plantarum JN01 were significantly lower (P < 0.05).

Huangjiu made with S. hirsuta J2 has improved metabolites and fermentation characteristics

Because of its highest activity of amylase, glucoamylase and protease, S. hirsuta J2 was selected for the subsequent fermentation experiment. Lab-scale (3 L) huangjiu fermentations were prepared with S. cerevisiae HJ and either S. hirsuta J2, A. flavus SU-16 (used in huangjiu production), A. oryzae MQ (used in sake production), or L. plantarum JN01 (used in sake and sherry production). Figure 4 shows the main physicochemical results of the huangjiu according to the national standards of China (GB/T 13662-2018). All the final fermentation products met the national standards except for huangjiu inoculated with L. plantarum JN01, Fermentations with S. hirsuta J2 had the highest fermentation rate (0 to 7.00% v/v in 25 h) and final alcohol content (16.03% v/v at 300 h) than fermentations inoculated with the other strains or the control group (traditional fermentation group) (P < 0.01) (Fig. 4A). Similarly, the reducing sugar content of the S. hirsuta J2 group remained low (<120 g/L) during the fermentation, and was significantly lower than that of the control group and those containing Aspergillus (Fig. 4B). Compared with the control group, the speed of A. flavus SU-16 and A. oryzae MQ groups fermentations were slightly higher, however the final alcohol contents (300 h) were no significant difference with the control group. At 300 h, sugar in A. flavus SU-16 and A. oryzae MQ groups was exhausted, and their values were not significantly different from these of the control groups. The fermentation inoculated with L. plantarum JN01 did not result in ethanol production (<2% v/v), but the reducing sugar content increased first and then decreased as the fermentation proceeds, and the maximum value is 114.06 ± 18.02 g/L at 50 h. The titratable acid content of the L. plantarum JN01 group also increased rapidly to 17.50 g/L, and the samples showed obvious rancidity (Fig. 4C). The titratable acid content of the other groups increased with fermentation, but remained within a reasonable and stable range. By the end of fermentation, the amino nitrogen content of each sample was in the 0.60–1.69 g/L range (Fig. 4D), which meets the superior standard of huangjiu18.

Fig. 4: Physicochemical parameters and main metabolites during the HJFM process.
figure 4

A Alcohol content (g/L); (B) Reducing sugar content (g/L); (C) Titratable acidity content (g/L); (D) Amino acid nitrogen content (g/L); (E) Heatmap of the main metabolites during the HJFM process. Clustering analysis was performed using the Pearson correlation coefficient and Euclidean distance based on the main flavor metabolites contents during the fermentation process. A value of 0.05 was set as the significance level; the data were marked as (*) p < 0.05, (**) p < 0.01, (***) p < 0.001, and (****) p < 0.0001. The p-value above 0.05 was considered as non-significant (ns). Error bars in histogram mead the standard deviation of observed data.

Full size image

The metabolites of huangjiu fermented with S. hirsuta J2, A. flavus SU-16, A. oryzae MQ, and L. plantarum JN01 and the control group were determined, and 28 main metabolites were identified. The different metabolites of huangjiu fermented with S. hirsuta J2 had similar proportions compared to those in the control group (Fig. 4E). Hierarchical cluster analysis showed similarities between S. hirsuta J2 and the control group based on their main metabolites. These results indicate that the metabolite profile of traditional huangjiu could be simulated by co-fermentation of S. hirsuta J2 and S. cerevisiae. By comparison, concentrations of metabolites were similar between huangjiu containing the A. flavus SU-16 and A. oryzae MQ inoculants, but not to the controls.

S. hirsuta J2 has the capacity to degrade BAs

The S. hirsuta J2 genome (Fig. 5A) encodes genes required for BAs degradation, including primary-amine oxidase (EC 1.4.3.21), acetyltransferase (EC 2.3.1.-) and spermidine synthase (EC 2.5.1.16) (Supplementary Table 4), according to the KEGG pathway analysis. These three enzymes bave the ability to degrade tyramine, cadaverine, phenylethylamine, putrescine, and histamine based on KEGG pathway analysis. This result indicates that S. hirsuta J2 has the potential to degrade BAs. Therefore, BAs production and degradation abilities of the S. hirsuta J2 was examined (Supplementary Fig. 4). It was found that the total content of BAs produced by S. hirsuta J2 was less than 2 mg/L, indicating that the strain basically did not produce BAs. The degradation rate of histamine, phenylethylamine and tyramine in the medium were more than 98%. The degradation rate of total BAs by S. hirsuta J2 was 72.30 ± 1.25%.

Fig. 5: Genome and fermentation features of S. hirsuta J2.
figure 5

Circular maps of S. hirsuta J2 genomes (A). From the outer circle to the inner circle: Circle 1, the location coordinates of the genome sequence of S. hirsuta J2; Circle 2, the GC content (71.42%); Circle 3, the GC skew value. Circle 4, annotation distribution of genome coding genes; Circle 5, annotation distribution of rRNA gene; Circle 6, annotation distribution of ncRNA gene; Circle 7, annotation distribution of tRNA gene; Circle 8, the genome size is 7.08 M. The BAs content and physicochemical parameters of huangjiu (B), sausage (C) and soy sauce (D) Error bars in histogram mead the standard deviation of observed data.

Full size image

S. hirsuta J2 reduces BAs content in huangjiu, soy sauce, and sausage

Before the fermentation was started, content of total BAs was determined to be 6.11 ± 0.32 mg/L in the HJFM (without cooked wheat qu). After the fermentation, the concentration of total BAs in huangjiu inoculated with S. hirsuta J2 (20.05 ± 0.43 mg/L) was significantly lower than huangjiu inoculated with A. flavus SU-16, A. oryzae MQ, and the control group (P < 0.01) (Fig. 5B). Histamine concentrations were reduced from 10.38 mg/L to 1.40 mg/L in huangjiu made with S. hirsuta J2, constituting an 86.51% reduction compared to the control group. Cadaverine and spermidine were also reduced by approximated 43.96% and 25.42%, respectively. By comparison, total BAs concentrations increased in huangjiu inoculated with A. flavus SU-16 from 36.48 ± 1.08 mg/L to 42.71 ± 1.16 mg/L compared to the control group. BAs content when A. oryzae MQ was added was indistinguishable from the control group.

To further explore the capacity of S. hirsuta J2 to degrade BAs in fermented foods, S. hirsuta J2 was inoculated into sausages and soy sauce for fermentation (Fig. 5C, D). The pH, moisture content, crude protein content, and soluble solid, total nitrogen and amino nitrogen content in soy sauce and sausage made with S. hirsuta J2 were not significantly different from the controls (P > 0.05). At the end of fermentation, BA concentration in the soy sauce and sausage of control group was 0.53 ± 0.06 g/100 mL and 229.03 ± 12.07 mg/kg, respectively. These values were lower when S. hirsuta J2 was added to the ferment, constituting a significantly decreased total BAs 43.41 ± 4.59% and 61.77 ± 6.22% compared with the control group in sausage and soy sauce fermentation, respectively.

Discussion

Fermented foods play a major role in the diets of many cultures worldwide. At present, fermented foods are increasingly recognized for their regional heritage19. Traditional fermented foods are usually produced by natural fermentation containing a multispecies community20. Generally, there are only limited number of known core species responsible for driving the fermentation process21. The transformation from natural fermentation to tractable fermentation with the core functional microorganism is essential for consistent quality of fermented foods. Although huangjiu can be made from other matrixes such as corn22,23, rice is the mostly used matrix for huangjiu fermentation, so the rice huangjiu is being studied here.

In this study, metagenome sequencing, metabolomics, strain isolation and characterization were combined to identify the core microbial taxa in huangjiu fermentations and the corresponding metabolites that they make. This was possible even though huangjiu is a highly complex food fermentations that has multiple fermentation steps and requires diverse bacterial and fungal species. Specifically, we showed that Saccharopolyspora, a genus of bacteria that has thus far only been associated with the production of bioactive natural products, is a key member of the huangjiu fermentation. The associations of Saccharopolyspora with metabolites was verified using S. hirsuta J2, a strain isolated from huangjiu. The utility of this strain was further confirmed by showing that it may be useful for reducing biogenic amines in huangjiu as well as other fermented food types.

Saccharopolyspora is the dominant member of the wheat qu and HJFM (huangjiu) microbiota. The genus Saccharopolyspora was first established by Lacey & Goodfellow24 and was assigned to the family Pseudonocardiaceae. Members of this genus are found in a wide range of habitats, including soil, marine sediments, marine invertebrates, plants and clinical samples. Secondary metabolites from Saccharopolyspora include the antibiotic erythromycin and the insecticides. Previous studies indicated that Saccharopolyspora was detected in vinegar25, nuruk26, daqu27,28, and huangjiu29. However, the importance of these bacteria in food fermentations was not determined. Based on O2PLS analysis, S. rectivirgula, S. erythraea, S. hirsuta, S. antimicrobica and S. cerevisiae were the primary contributors to the huangjiu metabolome. Among them, S. erythraea and S. hirsuta possessed the largest number of correlated metabolites (16 and 14 respectively), while S. cerevisiae was correlated with only 5 metabolites (ρ > 0.7, alcohols and esters). This indicated that Saccharopolyspora may be the main functional microorganism involved in the synthesis of metabolites in huangjiu fermentation, rather than S. cerevisiae or Aspergillus.

Our findings also showed that huangjiu fermentations (including wheat qu and HJFM) are dominated by bacteria, not fungi. This result is notable considering that huangjiu is an alcoholic fermentation, and A. flavus and S. cerevisiae is generally regarded as the main functional microorganisms in huangjiu fermentations13,30,31. A. flavus can produce a variety of extracellular enzymes such as α-amylase and glucoamylase that are responsible for the saccharification of rice. However, despite the known functional importance, A. flavus was of relative lower abundance, constituting only between 0.22% and 0.64% of the microbiota in HJFM. S. cereviseae is the traditional strain used to produce alcohol in the fermentation of alcoholic beverages32. In this study, as the fermentation progressed, the relative abundance of S. cerevisiae reaches up to 7.49% after the first stage (100 h), then decreased to 3.87% by stage II (300 h), which is consistent with its ethanol fermentation function. However, in addition to ethanol fermentation, S. cerevisiae from Shaoxing region was found to be mainly associated with the production of phenethyl alcohol and 3-methylthiopropanol30. Thus, the formation of the complex metabolites of huangjiu may originate from the metabolism of other microorganisms.

Based on this, we further explored the formation pathway of the main metabolites during huangjiu fermentation by mapping of the annotated genes identified in the metagenomes to the KEGG pathway. Our study found that Saccharopolyspora, Saccharomyces, and Aspergillus are the main participant in the synthesis of huangjiu metabolites. Moreover, correlation analysis suggested that Saccharopolyspora may be responsible for starch, cellulose and protein degradation during huangjiu fermentation, and not just the Aspergillus (like A. flavus, A. oryzae) as commonly thought33. Although A. flavus SU-16 has been used as the main resource of enzymes (protease, glucoamylase and liquifying enzyme etc.) in cooked wheat qu since 50 years ago. But additionally, Aspergillus may be involved only in the synthesis of Pro and Glu. Saccharomyces (S. cerevisiae and non-Saccharomyces yeast) may be involved in the synthesis of higher alcohols, esters, amino acids (Pro, Arg, Glu, Ala), and organic acids (acetate and citrate, lactate). Previous studies have found that non-S. yeast contribute to the production of esters and higher alcohols34. Also, in addition to being the main producer of lactate, it may be involved in the synthesis of higher alcohols.

Isolation and characterization of several Saccharopolyspora strains showed that these bacteria possess enzyme production (amylase, glucoamylase and protease) capacities. The utility of members of this genus was shown for S. hirsuta J2, an isolate from wheat qu. Inoculation of S. hirsuta J2 into HJFM could not only improve the rate of ethanol fermentation and final alcohol content, but also did not change the flavor of traditional huangjiu. There was no significant difference in fermentation parameters between the S. hirsuta J2 group and the control group (except alcohol content) at the end of fermentation. Importantly, the main metabolites of ferments containing S. hirsuta J2 were very similar to the control group. Additionally, the addition of S. hirsuta J2 resulted in a higher fermentation rate and alcohol content than control group (P < 0.0001). The increased fermentation rate may be due to improved breakdown of polysaccharides and proteins by amylases, glucoamylases, and acidic proteases produced by S. hirsuta J2 that result in more usable energy sources for S. cerevisiae.

BAs are a potential health risk substance and have been widely associated with food quality and safety3,35. In huangjiu, BAs are mainly produced in rice soaking and fermentation process16. There are no accurate regulations for BAs, but several countries including Germany, Belgium, France, and Switzerland have set regulations and limits for the upper limit of histamine, which are 2 mg/L, 6 mg/L, 8 mg/L and 10 mg/L respectively36. European Food Safety Authority (EFSA) recommended that the starter cultures lacking the capacity to produce BAs should be used when making fermented foods37. This is especially important in fermented food where high microbial activities may result in undesirable BAs accumulation38. Notably, S. hirsuta J2 can also degrade BAs. The total degradation rate of eight BAs was 72.30% in the culture medium, especially the six major BAs of tyramine (77.43%), putrescine (58.12%), histamine (99.55%), cadaverine (46.91%), phenylethylamine (98.19%) and tryptamine (98.41%). Although the capacity of S. hirsuta J2 to degrade BAs decreased slightly when added to complex fermentation environments, we found that compared with the control group, S. hirsuta J2 still degraded 45.05%, 43.41% and 61.77% of BAs content in huangjiu, soy sauce, and sausage. Notably, a different study showed that L. plantarum CAU 3823 isolated from huangjiu was able to degrade the BAs (40%)39. However, this value is slightly lower than S. hirsuta, suggesting the S. hirsuta J2 can adapt to different fermentation environments and can degrade BAs significantly.

By identifying the core functional organisms required to make fermented foods it may be possible to improve the scale and consistency of those product over time. Understanding the roles of microbial groups with important functions in various food fermentation systems has largely relied on bringing organisms into culture and assessing their metabolic capabilities. In recent years, metagenomics sequencing has been successfully used as an approach to connect taxonomy with these key fermentative functions and to assess the diversity of functional taxa in their fermentation systems40,41,42. Saccharopolyspora may be the potential functional strains leading the huangjiu fermentation. More studies on the application and function of Saccharopolyspora in food fermentation are required.

Methods

Wheat qu and HJFM sampling and pretreatment

Wheat qu and huangjiu fermentation mash were provided by a huangjiu factory (30°08′N, 120°49′E) in Shaoxing, Zhejiang, China on October 2018. As displayed in Fig. 1, a two-stage huangjiu fermentation process was performed in factory. The wheat qu was transported to the laboratory on dry ice within 24 h of collection and stored at −80 oC until analysis. In order to obtain adequate information and representation before carrying out analysis, 12 wheat qu blocks were selected randomly from upper, middle and lower locations in the storage room, all the blocks were ground to powder in a sterile grinder and mixed as a sample. The size of each block of wheat qu is approximately 23 × 15 × 7.5 cm (length × width × high), weighing around 2.5 kg. The HJFM was sampled in 1000 g quantities after 25, 50, 100 and 300 h of the production process in 30-ton fermentation tanks. Prior to sampling, the HJFM was mixed by aeration machines to improve sample homogeneity. Samples were transported on dry ice to the laboratory and immediately stored at −80 °C for metabolite and DNA analysis.

DNA isolation, quantitative real-time PCR (qPCR) and metagenome sequencing

Wheat qu and HJFM (5 g) were homogenized and grounded in liquid nitrogen, suspended in 10 mL of phosphate buffered saline buffer (PBS, pH 7.0), and then shaken for 5 min using a vortex adaptor. DNA was then extracted using sodium dodecyl sulfate and cetyl trimethyl ammonium bromide (SDS-CTAB)11. DNA concentrations were quantified using a UV-Vis Spectrophotometer Q5000 (Quawell, San Jose, USA). The DNA was stored at −20 °C until analysis. qPCR was used to detect the biomass of bacteria and fungi in wheat qu and HJFM according to the previous methods43. qPCR was performed using an qTOWER 3.0 system (Analytik Jena, Jena, Germany).

DNA libraries were constructed using 1 μg genomic DNA and the NEBNext® Ultra™ DNA Library Prep Kit (NEB, USA) following the manufacturer’s recommendations. Briefly, DNA was fragmented by sonication to a size of 350 bp, and then the DNA fragments were end-polished, A-tailed, and PCR was used to ligate a full-length adaptor for Illumina sequencing. The libraries were purified (AMPure XP system), evaluated for size distribution on an Agilent 2100 Bioanalyzer (Agilent Technologies), and quantified using real-time PCR according to the qPCR Quantification Protocol Guide (Agilent Technologies). The Illumina platform (Illumina Hiseq 4000) with 2 × 150 bp paired-end sequencing was used for sequencing each DNA library (Novogene, China).

Metabolites profiling

To investigate relationship between metabolites and microbial community and function, metabolites in wheat qu and HJFM were measured. Wheat qu solid powder (10 g) was mixed with sterile water (20 mL) by rotational shaking at 150 rpm for 2 h at room temperature and then filtered through a double layer of filter paper (Whatman International, Ltd., England). The HJFM samples were also filtered through a double layer of filter paper. Both filtrates were used for the following experiments.

(1) Organic acids

Reverse-phase high-performance liquid chromatography (RP-HPLC) was used to determine quantities of the eight organic acids: α-ketoglutaric acid, pyruvate, oxalate, acetate, malate, tartrate, citrate, and lactate. Wheat qu and HFJM filtrates were mixed with 2 mL of trichloroacetic acid (TCA, 10%, w/v), kept at 4 °C overnight, and then centrifuged at 8000 g for 15 min. The supernatants were then analyzed using an RP-HPLC system (Waters e2695, Milford, MA) equipped with an Athena C18-WP column (250 × 4.6 mm; 5 μm) and a UV detector. The column temperature was maintained at 30 °C. The detection wavelength was 210 nm. The mobile phase was phosphate buffer (0.025 M NaH2PO4, pH 3.1), and the flow rate was 0.7 mL/min.

(2) Free amino acids. Wheat qu and HJFM were analyzed for the presence of aspartate (Asp), glutamate (Glu), serine (Ser), histidine (His), glycine (Gly), threonine (Thr), arginine (Arg), alanine (Ala), tyrosine (Tyr), cysteine (Cys), valine (Val), methionine (Met), phenylalanine (Phe), isoleucine (Ile), leucine (Leu), lysine (Lys), and proline (Pro) according to the previous methods44. Briefly, the filtrates were mixed with 2 mL of TCA (10%, w/v) for 4 h, and then centrifuged at 8000 g for 30 min. The supernatant was then analyzed using an RP-HPLC system equipped with an ODS HYPERSIL column (250 × 4.6 mm, 5 μm) and a UV detector. The column temperature was maintained at 40 °C. The detection wavelengths were 338 nm and 262 nm, and the flow rate was 1.0 mL/min.

(3) Volatile compounds. The volatile compounds were determined by using headspace solid-phase microextraction (HS-SPME) and analyzed using gas chromatography-mass spectrometry (GC-MS) according to the method reported with modifications45,46,47,48. The filtrates (5 mL), prepared as described above was placed in a 20-mL glass SPME vial together with 3.0 g of sodium chloride and 25 µL of 2-octanol (41.2 mg/L in absolute ethanol) as an internal standard. The vial was incubated for 50 min at 50 °C. After extraction, the fiber was introduced into the injection port of the GC-MS system (at 250 °C for 7 min) and the analytes extracted from the fiber were thermally desorbed. Mass spectra and retention indices of compounds detected by GC-MS analysis were compared with those found in published data and the MS library of the National Institute for Standards and Technology (NIST, Search 2.0). The quantity of volatile compounds was established by using the internal standard method and external standard calibration curve methods.

Bioinformatic analysis

Sequence reads were first screened to remove low-quality reads (quality value less than 38 ≥ 40 bp, and ≥10% N containing reads) by Readfq (V8, https://github.com/cjfields/readfq). DNA sequence reads that aligned to the wheat, rice and human genomes were removed (Bowtie 2.2.4, parameters: --end-to-end, --sensitive, -I 200, -X 400). The remaining high-quality reads were assembled and analyzed using SOAP denovo software version 2.04 (parameters: -d 1, -M 3, -R, -u, -F, -K 55)49,50,51,52,53. Unused reads from each sample were assembled using the same parameters. Genes were predicted using MetaGeneMark (Version 2.10) software with default parameters54,55,56,57,58. This process produced 14.92 Gbp of raw data per sample on average. Duplicate, low quality and host reads were removed from the raw data set. For taxonomic assignment, the predicted genes were aligned to the integrated non-redundant database (NR) using DIAMOND software (Version 0.9.9) with E ≤ 1e-559. For functional analysis, all the genes in our catalogue were aligned to the KEGG database (http://www.kegg.jp/kegg/) using DIAMOND (Version 0.9.9, default parameter except that -k 50 –sensitive -e 0.00001).

To obtain statistics of the relative abundance of different functional hierarchies, the relative abundance of each functional hierarchy was considered equal to the sum of the relative abundances annotated to that functional level. To elucidate the biosynthetic pathways responsible for the metabolites identified within huangjiu fermentation samples, the gene ID numbers of related enzymes, the relative abundance of corresponding enzymes, and the taxonomy annotation were extracted from the KEGG annotation. Metabolic pathways of functional microorganism were constructed by using KEGG annotations. DNA sequencing, gene prediction, assembly and annotation statistics are shown in Supplementary Table 5.

Isolation of Saccharopolyspora and whole-genome sequencing of S. hirsuta J2

For microbial isolation on laboratory culture medium, wheat qu samples were suspended in sterile distilled water, and serial dilutions were plated onto Gauzeˊs Medium (Sigma) for incubation at 30 °C for between 7 to 14 days. Since members of Saccharopolyspora are predominant in wheat qu and HJFM, and were likely the core functional microorganisms in HJFM, especially Saccharopolyspora hirsute, Saccharopolyspora-like colonies60 were isolated, purified, and preserved in 30% w/v glycerol in water at −80 °C. 16 S rRNA gene sequencing using the 16S-27F and 16S-1492 primers was used for taxonomic identification61. The isolated strains were identified as three S. hirsuta (J1 and J3), two S. rectivirgula (J4 and J5), three S. shandongensis (J6 to J8), and one S. erythraea (J9). S. hirsuta J2 was deposited in China Center for Type Culture Collection under the number CCTCC M 2020103. The whole genome of S. hirsute J2, the strain with the highest activity of amylase, glucoamylase and protease, was sequenced using Nanopore PromethION platform (Oxford Nanopore Technologies Ltd, Oxford, UK) and Illumina Navaseq PE150 (Illumina, SanDiego, CA, USA) at the Beijing Novogene Bioinformatics Technology Co., Ltd.

Preparation and enzyme activities of cooked wheat qu

To analyze the enzyme activities of wheat qu, that of samples from factory and artificially cooked wheat qu by isolated microbial strains were compared. The wheat qu sampled from factory was used as the control, A. flavus SU-16 cooked wheat qu, A. oryzae MQ cooked wheat qu and Lactiplantibacillus plantarum JN01 cooked wheat qu were used as references, and nine cooked wheat qu by Saccharopolyspora species J1 to J9 respectively were the experimental group.

To prepare the cooked wheat qu inoculation, wheat was prepared by crushing in a barley press roller and sterilization at 121 °C for 30 min. After cooling to <40 °C, strains of selected A. flavus SU-16, A. oryzae MQ, L. plantarum JN01, and Saccharopolyspora species J1 to J9 were inoculated (2.0 × 106 cfu/g) into the sterilized wheat separately and stirred until homogeneous with additional water approximately 25% of total weight. The inoculated samples were incubated for 72 h at 28 °C (A. flavus SU-16 and A. oryzae MQ) or 37 °C (9 isolated Saccharopolyspora species and L. plantarum JN01). The final (also known as cooked) wheat qu was dried until its moisture content reached less than approximately 15%.

Protease, glucoamylase and amylase activities in wheat qu sampled from the brewery and 12 cooked wheat qu were determined as previously described11. One unit of amylase activity was defined as the amount of dry sample required for the liquefaction of 1 g starch per hour in sodium phosphate buffer (100 mM, pH 4.6) at 30 °C (U/g dry sample). One unit of glucoamylase activity was defined as the amount of dry samples required for the liberation of 1 mg glucose per hour in sodium acetate buffer (50 mM, pH 4.6) at 30 °C (U/g dry sample). One unit of protease activity was defined as the amount of dry samples required for the liberation of 1 μg of tyrosine per min in sodium acetate buffer (50 mM, pH 4.6) at 30 °C (U/g dry sample).

Characterization of BAs metabolism by S. hirsuta J2

To evaluate the capacity of S. hirsuta J2 to produce BAs, that strain was inoculated at a level of 1 × 107 cells/mL in Gauzeˊs liquid medium (Guangdong Huankai Microbial Sci. & Tech.co., Ltd, China) containing 1.0 g/L of either histidine, tyrosine, tryptophan, phenylalanine, ornithine monohydrochloride, lysine, or agmatine sulfate salt and incubated at 30 °C for 24 h. All amino acids were purchased from Sigma. Negative controls were not inoculated. BAs quantification was performed as previously described62 using an RP-HPLC system (Waters e2695, Milford, MA) equipped with an XBridge C18 (250 mm × 4.6 mm, 5 µm) and UV detector.

For quantification of the capacity of S. hirsuta J2 to consume BAs, S. hirsuta J2 was inoculated into Gauzeˊs liquid medium containing 50 mg/L each of histamine, tyramine, cadaverine, putrescine, spermine, spermidine, tryptamine, and phenylethylamine (each from Sigma). All cultures, including the negative control, were incubated at 30 °C for 5 days. BA consumption was measured by collecting S. hirsuta by centrifugation at 10,000 g for 10 min and then using the supernatant to measure BA concentrations by HPLC. S. hirsuta J2 BA degradation was calculated according to the formula:

$${\text{BAs degradation}} (\%) = [(A-B)/A]\times 100\%$$

A was the concentration of BA in the negative controls and B was the concentration of the BA in the medium incubated with S. hirsuta J2.

Application of S. hirsuta J2 in huangjiu, sausage and soy sauce fermentations

To evaluate the capability of degrading BAs in fermentation, S. hirsuta J2 and several other strains were used for fermentation of Huangjiu, sausage, and soy sauce. Notably, all the fermentation microorganisms used here do not produce biogenic amines, and biogenic amines producing in huangjiu fermentation is spontaneous.

(1) Huangjiu fermentation. Lab-scale huangjiu fermentations (3 L) were prepared in triplicate (n = 3) with S. cerevisiae HJ63 and either S. hirsuta J2, A. flavus SU-169, Aspergillus oryzae MQ9, or Lactiplantibacillus plantarum JN0116 cooked wheat qu. Huangjiu was produced with the techniques used by a huangjiu brewery in Shaoxing (Zhejiang Province), China (Fig. 1). In brief, after the steamed rice was cooled to <30 °C, fermentation was initiated by adding cooked wheat qu (13.6 g/100 g uncooked rice), S. cerevisiae HJ (11.4 mL/100 g uncooked rice), and water (125 mL/100 g uncooked rice). Wheat qu sampled from the brewery was used as a positive control (the control group). The rice mixtures were incubated at 30 °C for 100 h and then maintained at 16 °C for 200 h. Huangjiu was examined for reducing sugars using the 3,5-dinitrosalicylic acid (DNS) method64. Titratable acidity, alcohol content, and amino acid nitrogen were measured using standard methods65. BAs, amino acids, organic acids, and volatile compounds were quantified as described before.

(2) Sausage fermentations. Sausage fermentations were performed in triplicate (n = 3) in laboratory following the method described by Li et al. (2019). In short, fresh pork was bought in local supermarket in Wuxi. Visible fat and connective tissues were removed from commercially purchased pork (lean meat (80%) and back fat (20%)). The meat was ground and then mixed with 0.1% (w/w) spice powder (spices, garlic, chilli, pepper, cardamom), 0.01% (w/w) sodium nitrite, 2% (w/w) salt, 0.2% (w/w) compound phosphate, 0.05% (w/w) sodium ascorbate. Sausage fermentation was spontaneous, so no microbial inoculant was used. Sausages were manufactured with S. hirsuta J2 (106 cfu/g). Sausages without added S. hirsuta J2 was used as a control group. The sausages were first incubated for one day at 30 ± 0.5 °C and a relative humidity (RH) of 90%. The sausages were then incubated at 16 ± 0.5 °C with successive 24 h passages at a RH of 90%, 80% and the 75% for drying. The sausages were transferred to 14 ± 0.5 °C and a RH of between 65 to 70% for 9 days. Each sausage type was made in triplicate (n = 3).

(3) Soy sauce fermentation. Soy sauce was prepared in triplicate (n = 3) using the method described by Yang et al.66, with slight modifications. As above, dry soybeans were bought in local supermarket in Wuxi, and these with low quality were filtered. Cooked soybeans were mixed with roasted wheat (7:3 ratio) and then fermented for 72 h at 35 °C with A. oryzae (106 cfu/g) or S. hirsuta J2 (106 cfu/g) to prepare the koji, a pre-fermented products used to add flavor in fermentation. Soy sauce added A. oryzae was used as a control group. The koji was mixed with a brine solution (20% w/v NaCl) and fermented for 40 days at between 40 to 45 °C. The soy sauce was then prepared by pressing, filtration and pasteurization.

Statistical analyses

The R packages ggplot2 and pheatmap were used for data visualization. A bidirectional orthogonal partial least squares (O2PLS) model was used to elucidate the association between microbial species and metabolites produced during huangjiu fermentation. O2PLS analysis was performed using SIMCA 14.1 software (Umetrics AB, Umeå, Sweden). Differences between the mean values for individual groups were assessed with variance tests (two-way analysis of variance (ANOVA) and one-way ANOVA) and Duncan’s multiple range tests. A value of 0.05 was set as the significance level; the data were marked as (*) P < 0.05, (**) P < 0.01, (***) P < 0.001, and (****) P < 0.0001. The p-value above 0.05 was considered as non-significant (ns).