甘肃农业大学：食品科学与工程学院（文献讲义）High-throughput sequencing approach to characterize dynamic changes of the fungal and bacterial communities during the production of sufu, a traditional Chinese fermented soybean food.pdf

Contents lists available at ScienceDirect Food Microbiology journal homepage: www.elsevier.com/locate/fm High-throughput sequencing approach to characterize dynamic changes of the fungal and bacterial communities during the production of sufu, a traditional Chinese fermented soybean food Dandan Xua,b , Peng Wanga,b , Xin Zhanga , Jian Zhanga,b , Yong Suna , Lihua Gaoa , Wenping Wanga,∗ a Beijing Academy of Food Sciences, 100068, Beijing, China b Beijing Food Brewing Institute, 100050, Beijing, China ARTICLE INFO Keywords: Sufu Fermentation Fungal communities Bacterial communities High-throughput sequencing ABSTRACT Red sufu is a traditional food produced by the fermentation of soybean. In this study, sufu samples were periodically collected during the whole fermentation to investigate the dynamic changes of fungal and bacterial communities using high-throughput sequencing technology. The overall process can be divided into pre- and post-fermentation. During post-fermentation, the pH value showed a gradual decrease over time while the amino nitrogen content increased. Trichosporon, Actinomucor and Cryptococcus were the main genera in pre-fermentation while Monascus and Aspergillus were dominant in post-fermentation. This huge shift in fungal composition was caused by process procedure of pouring dressing mixture. However, the bacterial composition was not greatly changed after pouring dressing mixture, the Acinetobacter and Enterobacter were the predominant genera throughout the whole process. Furthermore, Bacillus species were first detected after adding dressing mixture, but declined abruptly to a very low level (0.07%) by the end of the fermentation. Our work demonstrates the dynamic changes of physicochemical properties and microbial composition in every fermentation stage, the knowledge of which could potentially serve as a foundation for improving the safety and quality of sufu in the future. 1. Introduction Sufu is a soft, cheese-like traditional food produced by the microbial fermentation of soybean (Yang et al., 2014b). In China it has been widely consumed as an appetizer and a side dish since the Wei Dynasty (220–265 CE) (Chung et al., 2005). Since it is made from fermented tofu, which is a good protein and calcium source, sufu is known as a healthy and low cholesterol food of plant origin (Qiu et al., 2018). It is becoming increasingly popular in Asia because of its unique flavor and taste. On the basis of the different microbial starter culture, sufu can be classified into mould-fermented and bacteria-fermented (Han et al., 2004). The mould-fermented sufu is the most dominant type among all sufu types (Han et al., 2003). The red sufu is a typical mould-fermented sufu in China. It is manufactured by first cultivating a fungus such as Actinomucor, Mucor, or Rhizopus on the surface of tofu cubes to prepare the pehtze. The prepared pehtze is salted in water for about 5 days, during which time the pehtze absorbs salt. The high concentration of salt can inhibit the continued growth of mould and reduce the contamination of bacteria in the environment while imparting salty taste to the pehtze. Then the salt-pehtze were carefully placed in the sterilized glass bottle, the dressing mixture was subsequently poured into it. The salt-pehtze ripens in the dressing mixture in a time period of 3 months. Various enzymes secreted by the microorganisms decompose the raw materials into alcohols, aldehydes, organic acids, esters, and amino acids through a series of biochemical processes thereby imparting a pleasant taste. Until now, studies in sufu are mainly focused on its biochemical and physiological properties, antioxidant activity, safety assessment, volatile components etc. (Cai et al., 2016; Chen et al., 2012; Huang et al., 2011b; Ma et al., 2013b; Moy and Chou, 2010; Moy et al., 2012; Xia et al., 2014). It has been shown that the microbial community plays an important role during the long period of food fermentation (Song et al., 2017). However, studies investigating the microbial dynamic changes in the fermentation of sufu are limited. Yan et al. analyzed the mesophilic aerobic bacteria (TMAB), lactic acid bacteria (LAB), and fungal properties during sufu manufacturing using the flat colony counting method, and found that TMAB, LAB, and fungal counts were low in https://doi.org/10.1016/j.fm.2019.103340 Received 29 January 2019; Received in revised form 19 September 2019; Accepted 20 September 2019 ∗ Corresponding author. Beijing Academy of Food Sciences, No. 70. Yangqiao Road, Beijing, 100068, China. E-mail address: wwpsmn@163.com (W. Wang). Food Microbiology 86 (2020) 103340 Available online 21 September 2019 0740-0020/ © 2019 Elsevier Ltd. All rights reserved. T

tofu. After fermentation, an increased microbial count in the pehtze was observed. A significant decline in microbial count occurred after salting, and almost no fungal growth (< 1 log cfu/g) could be detected (Ma et al., 2013a). Feng et al. evaluated the bacterial flora during the ripening of Kedong sufu (a typical bacteria-fermented sufu) using polymerase chain reaction denaturing gradient gel electrophoresis (PCR-DGGE) and culturing methods. They found that Enterococcus avium, Enterococcus faecalis, and Staphylococcus carnosus were the dominant microflora throughout the fermentation of sufu (Feng et al., 2013). Since the results from other studies are only for microorganisms that can be easily cultivated, methods that rely on culture have been insufficient to fully understand the microbial population. PCR-DGGE is a common method for evaluating the composition of microorganisms, but it is time consuming and has limited ability to detect rare or uncultivable microorganisms (Hong et al., 2016). High-throughput sequencing has been widely used to characterize the composition of the microbial community of fermented food, such as wine, vinegar, soy sauce, etc. (Sulaiman et al., 2014; Tang et al., 2017; Wang et al., 2016). Based on this technology, the aims of this study were as follows: 1) investigate the dynamic changes of physicochemical properties and fungal structure during sufu fermentation and their correlation with process procedures, 2) identify the relative abundance and diversity of bacteria taxa in sufu samples, which is crucial for the flavor and security of fermented food. 2. Materials and methods 2.1. Sufu preparation and sample collection Red sufu samples were prepared and collected at the Wangzhihe Food Co. Ltd. A diagram of the production and sampling points with sample names are presented in Fig. 1. Simply, pehtze is prepared by inoculating Actinomucor elegans on the surface of tofu cubes, then salting the pehtze for about 5 days, and dispense the salt-pehtze into wide-mouthed glass bottles and ripens in the dressing mixture for a period of 3 months. The dressing mixture of red sufu mainly consists of red kojic rice (cooked rice inoculated with Monascus purpureus), edible alcohol, sugar, chiang (flour paste fermented by Aspergilus oryzae) and spices. Sufu samples were collected periodically at 9 different stages of fermentation: tofu (T), pehtze which inoculated with A. elegans for 24 h (A24), 48 h (A48), salt-pehtze (S), fermentation of sufu for 0 day (D0), 5 days (D5), 1 month (M1), 2 months (M2), 3 months (M3). Samples were collected from five independent batches and used as replicates. A total of 45 samples were transported into the lab on ice and stored at −80 °C until further use. 2.2. Physicochemical analysis Five grams of the sufu samples were homogenized with 50 mL of distilled water followed by pH measurements using a pH meter (Mettler Toledo). The amino nitrogen content and NaCl concentration were determined according to SB/T10170-2007 standard. Sufu samples (20 g) were boiled with distilled water (80 mL) with gentle stirring. Boiled sufu slurry was diluted to 200 mL with distilled water. The sufu solution was filtered with dry filter paper and then the filtrate was used to measure amino nitrogen content and NaCl concentration. 10 mL of the filtrate was mixed with 50 mL water and titrated to pH 8.2 with 0.05 mol/L NaOH and then 10 mL of 36% (w/v) formalin solution was added. The mixture was titrated to pH 9.2 with 0.05 mol/L NaOH. The volume of consumed NaOH for raising pH (from 8.2 to 9.2) was recorded to determine amino nitrogen content. To determine the NaCl concentration, 2 mL of the filtrate was mixed with 50 mL water and titrated with 0.100 mol/L AgNO3 using 5% (w/v) K2CrO4 solution (1 mL) as an indicator. The titration was terminated when the solution appeared orange. The content of reducing sugar was determined according to previous study (Van Waes et al., 1998). Samples from the same stage of the five independent batches were mixed for physicochemical analysis. All tests were carried out in triplicate. 2.3. DNA extraction Five grams of sufu samples were mixed with 25 mL 0.1 mol/L TrisHCl (pH 8.0), shaken well, and filtered through three layers of sterile gauze. The filtrate was centrifuged at 10,000×g for 20 min at 4 °C. The pellets were used for DNA extraction by GMO food DNA Extraction Kit (Tiangen, Beijing, China) following the manufacturer's protocol. The total DNA concentration and quality were checked using a NanoDrop 2000 (Thermo) spectrophotometer and agarose gel electrophoresis. 2.4. 16S rRNA gene amplicon sequencing and ITS amplicon sequencing Variable regions V3–V4 on microbial 16S rRNA gene of bacteria and the ITS2 region of fungi were amplified using PCR (95 °C for 3 min, followed by 30 cycles at 98 °C for 20 s, 58 °C for 15s, 72 °C for 20 s and a final extension at 72 °C for 5 min). The microbial 16S rRNA gene were amplified by forward primer F341 5′- ACTCCTACGGGRSGCAGCAG -3′ and reverse primer R806 5′- GGACTACVVGGGTATCTAATC -3′ (Klindworth et al., 2013). ITS2 were amplified with forward primer F2045 5′-GCATCGATGAAGAACGCAGC-3′ and reverse primer R2390 5′-TCCTCCGCTTATTGATATGC-3′ (Hirokazu et al., 2012). PCR reactions were performed in 30 μL mixture containing 15 μL of 2 × KAPA Fig. 1. The production diagram of sufu used in this study with the sampling points indicated. Tofu (T), pehtze which inoculated with A. elegans for 24 h (A24), 48 h (A48), salt-pehtze (S), fermentation of sufu for 0 day (D0), 5 days (D5), 1 month (M1), 2 months (M2), 3 months (M3). D. Xu, et al. Food Microbiology 86 (2020) 103340 2

Library Amplification ReadyMix, 1 μL of each primer (10 μmol/L), 10 ng of template DNA and ddH2O. PCR products were detected using 2% agarose gel electrophoresis, and then purified by AxyPrep DNA gel Extraction Kit (AXYGEN). Amplicon libraries were quantified using a Qubit 2.0 Fluorometer (Thermo Fisher Scientific). Amplicons were then sequenced using Illumina HiSeq PE250 at Realbio Genomics Institute (Shanghai, China). 2.5. Bioinformatics analysis The paired-end reads were assembled into longer tags after sequencing and then quality-filtered. Tags were restricted between 220 bp and 500 bp and the average Phred score of bases was not < 20 (Q20) and not > 3 ambiguous N. Bioinformatic analyses were performed using QIIME (v1.7.0) on the extracted high-quality sequences (Caporaso et al., 2010). First, the sequences were aligned using PyNAST (Caporaso et al., 2009) and clustered under 100% sequence identity using UCLUST (Edgar, 2010) to obtain the unique sequence set and clustered into operational taxonomic units (OTUs). Then these representative sequences were further classified into operational taxonomic units (OTUs) with a 97% similarity using UCLUST. Each representative sequence was assigned to a taxa by Ribosomal Database Project, (RDP, Release 11.5) (Cole et al., 2006) and Greengenes (Release 13.8) (DeSantis et al., 2006) to obtain the taxonomy information of phylum, class, order, family, genus and species. 2.6. Statistics analysis Differences in the relative abundances of taxonomic levels between samples were evaluated using the Mann-Whitney test. Values of P < 0.05 were considered significantly different between different groups. Alpha diversity (Shannon index) analyses were analyzed using QIIME. A de novo taxonomic tree was constructed using FastTree (Price et al., 2009) for alpha and beta diversity calculations. To evaluate alpha diversity, the Shannon Wiener and the Chao1 were calculated. UniFrac metrics were calculated to evaluate beta diversity (Lozupone and Knight, 2005). Both weighted and unweighted principal coordinate analysis (PCoA) were performed. Redundancy analysis (RDA) was performed using CANOCO (canonical community ordination) 4.5 software. The graphic presentations were generated by the R package version 3.1.2 and the origin software package version 8.5. Significant differences of physicochemical properties were calculated with the Duncan's multiple range test using SAS (version 9.1.2) and the graphic presentations were generated by GraphPad prism 7. Differences were considered statistically significant when P < 0.05. 3. Results 3.1. Physicochemical properties of the sufu The physicochemical properties were mainly affected by the production procedures and fermentation time. The initial pH was 6.07 which significantly increased to 6.39 at salt-pehtze stage (P < 0.05). From the D0 until the end, the pH value showed a gradual decrease (from 6.44% to 6.33%) (Fig. 2A). The NaCl concentration peaked at the salt-pehtze stage (13.6%) and decreased significantly after adding dressing mixture (8.02%) (P < 0.05). Concentrations of NaCl continued to drop until M1 (6.43%) and then increased to 8.30% during the following stages (Fig. 2B). The amino nitrogen content peaked at the end of the pehtze stage (0.65%) and decreased significantly to 0.26% at D0 (P < 0.05). Thereafter, the amino nitrogen continued to rise from 0.26% to 0.59% during the ripening fermentation stage (Fig. 2C). Reducing sugar was not detected before pouring dressing mixture. It peaked at M1 (10.20%) and was significantly higher than that of other stages during the post-fermentation (P < 0.05) (Fig. 2D). 3.2. ITS2 rRNA gene sequencing of the fungal community in sufu The Illumina Hiseq platform was employed to sequence the internal transcribed spacer 2 (ITS2) of 45 samples, thereby obtaining 1.55 million clean tags (34,606 ± 5201 on average) after concatenation and quality control. The clean reads with an average length of 335 bp were clustered into 334 OTUs at 97% similarity level. These OTUs were clustered into 3 phyla and 139 genera. Ascomycota (accounting for 5.13%–46.15%), Basidiomycota (accounting for 16.09%–55.79%) and Zygomycota (accounting for 9.61%–78.77%) were the main phyla during the pehtze and salt-pehtze stage. Then Ascomycota became the dominant phyla that accounts for 90.27%–99.42% in the following stages (Fig. 3A). In the genus level, Trichosporon (29.4%) at A24 and Actinomucor (78.8%) at A48 were the most abundant respectively. The Cryptococcus (40.0%) and Actinomucor (29.8%) were the most abundant genera at salt-pehtze stage. But after pouring dressing mixture, Monascus and Aspergillus became the main genera during the ripening fermentation stage. Monascus increased from D0 (36.87%) to M1 (70.42%) and at the same time Aspergillus declined from 32.12% to 6.22%. Meyerozyma, Millerozyma and Pichia were also the main genera at D0 and D5. Interestingly, at the end of ripening fermentation stage (M3), Monascus (46.91%) and Aspergillus (40.00%) were at a similar ratio as they were at D0 - the beginning of ripening fermentation stage (the Aspergillus/Monascus ratio was 0.87 at D0 and 0.85 at M3) (Fig. 3B). The similarity in analysis among different sufu samples based on the relative abundance of fungal OTUs showed that besides T (not included in the fermentation period), the samples could be organized into two main branches. The first branch was composed of samples of three time points, A24, A48 and S. The second branch was composed of samples of five time points: D0, D5, M1, M2 and M3. These changes were in accord with the production procedures. Since the sufu was mainly fermented by fungus, we can divide the whole fermentation process into pre-fermentation and post-fermentation by before or after pouring the dressing mixture (Fig. 4). 3.3. 16S rRNA gene sequencing of the bacterial community in sufu The Illumina HiSeq platform was employed to sequence the V3–V4 regions of 16S rRNA gene of 45 samples, thereby obtaining more than 1.57 million clean tags (35,049 ± 1721 on average) after concatenation and quality control. The clean reads with an average length of 427 bp were clustered into 22,466 OTUs at 97% similarity level (Information of the sample statistic has been listed in the Supplementary Table S1). These OTUs were clustered into 20 phyla, 47 classes, 71 orders, 153 families, and 355 genera. In terms of the relative abundance, Proteobacteria, Firmicutes, and Bacteroidetes were the dominant phyla throughout all the stages of the fermentation (Fig. 5A). Proteobacteria were the most abundant phylum at every stage; especially at the end of the fermentation where its relative abundance was 90.65%. Firmicutes were much less abundant in the salt-pehtze (7.88%) than in the pehtze (32.71%) stages. Contrastingly, Bacteroidetes were much more abundant in the salt-pehtze (24.24%) than in the pehtze (10.81%) stages. At the end of the fermentation, the relative abundance of Firmicutes and Bacteroidetes declined to 4.21% and 4.26%, respectively. At the genus level, Acinetobacter were the most abundant classified genus from the pehtze stage to M2 (ranging from 11.90% to 51.46%) and the second most abundant at M3 (11.46%). The relative abundance of Enterobacter was < 4% from A24 to D5, but dramatically increased to 24.66% at M1. Contrastingly, the relative abundance of Empedobacter was > 7% from A24 to D5, but dramatically declined to 0.64% at M1. Meanwhile Lactococcus declined rapidly from M1 to M3. Streptococcus and Weissella were also the main genera during the pehtze stage, but they became less abundant in the following stages. Pseudomonas and Klebsiella were more abundant in the ripening fermentation stage than the pehtze stage. Bacillus was not detected during the pre-fermentation stage (A24, A48, and S). D. Xu, et al. Food Microbiology 86 (2020) 103340 3

However, it was spotted at D0 which increased significantly from D5 to M2. The abundance of Bacillus declined rapidly to an extremely low abundance (0.07%) at the end of the ripening fermentation stage (Fig. 5B). The Shannon index showed that the diversity of the bacterial communities increased slightly at D0, then showed no significant differences in M1 and M2, but declined in M3 (Supplementary Table S1), mainly because of the dominance of Acinetobacter. The structure of the bacterial communities in the samples from the 9 different stages was also compared using weighted and unweighted UniFrac PCoA. Although some overlap was presented in the score plots among the samples from the 9 stages, the data points were largely separated in both weighted (accounting for 35.12% and 18.35% of the total variance by the first 2 principal components (PC), respectively; Fig. S1A) and unweighted (accounting for 14.00% and 10.90% of the total variance by the first 2 PC, respectively; Fig. S1B) analyses. By weighted analysis, the samples from stage T was obviously separated from all the other stages, and only M3 was separated from the other stages. By unweighted analysis, the samples from stage T were also separated from all the other stages. Except stage T, M2 and M3 were significantly clustered separately and different from the other stages. A further multivariate ANOVA test showed significant differences in the bacterial communities from the 9 different stages based on the weighted (P < 0.05) UniFrac analysis (Fig. S2A). The M2 and M3 stages were significantly different from other stages. The pre-fermentation stage could significantly separate from the post-fermentation. For the unweighted (P < 0.05) UniFrac analysis, compared to other groups, the most significant difference in bacterial composition was found in samples from stage T (Fig. S2B). According to the bioplot of RDA in S and M3 (Fig. 6A), three different bacteria represented the main microbial changes of the two timespots. Acinetobacter was observed in the left part of the plot, indicating that it was much more abundant in S than M3. However, Qingshengfania and Propionibacterium were observed in the other side of the plot, indicating that these bacteria were more abundant in M3 than S. Fig. 6B shows that different bacteria represented two different time points which is a similar observation to Fig. 6A. Raoultella, Klebsiella, and Empedobacter were spotted in the left side of plot, showing they were more abundant in A48 than S. On the right side, Streptococcus and Weissella were observed, showing they can represent S by having a higher abundance than A48. 4. Discussion Sufu is a typical, traditional Chinese fermented soybean curd; it is Fig. 2. Physicochemical changes in sufu. (A) pH mean values at various fermentation stages. The initial pH was 6.07 which significantly increased to 6.39 at saltpehtze stage (P < 0.05). From D0 to M3, the pH value showed a gradual decrease. (B) NaCl concentration at various fermentation stages. The NaCl concentration peaked at the salt-pehtze stage (13.6%) and decreased significantly at D0 (8.02%) (P < 0.05). (C) Amino nitrogen content at various fermentation stages. The amino nitrogen content peaked at the end of the pehtze stage (0.65%) and decreased significantly to 0.26% at D0 (P < 0.05). Thereafter, the amino nitrogen continued to rise from 0.26% to 0.59% during the ripening fermentation stage. (D) Reducing sugar at various fermentation stages. Reducing sugar was not detected before pouring dressing mixture. It peaked at M1 (10.20%) and was significantly higher than that of other stages during the post-fermentation (P < 0.05). Different letters indicate significant differences (P < 0.05). Tofu (T), pehtze which inoculated with A. elegans for 24 h (A24), 48 h (A48), salt-pehtze (S), fermentation of sufu for 0 day (D0), 5 days (D5), 1 month (M1), 2 months (M2), 3 months (M3). D. Xu, et al. Food Microbiology 86 (2020) 103340 4