Integrated Meta-omics Approaches To Understand the Microbiome of Spontaneous Fermentation of Traditional Chinese Pu-erh Tea

Fermented foods play important roles in diets worldwide and account for approximately one-third of all foods and beverages consumed. To date, traditional fermentation has used spontaneous fermentation. The microbiome in fermentation has direct impacts on the quality and safety of fermented foods and contributes to the preservation of traditional methods. Here, we used an integrated meta-omics approach to study the microbiome in the fermentation of pu-erh tea, which is a well-known Chinese fermented food with a special flavor and healthful benefits. This study advanced the knowledge of microbiota, metabolites, and enzymes in the fermentation of pu-erh tea. These novel insights shed light onto the complex microbiome in pu-erh fermentation and highlight the power of integrated meta-omics approaches in understanding the microbiome in food fermentation ecosystems.

fermentation. In addition, little is known about the enzymes involved in the metabolism of phenolic compounds, which are essential for tea processing.
To conduct a systematic review of the fermentation mechanism of pu-erh tea, the microbiome in fermentation was studied using an integrated meta-omics approach. Bacterial 16S rRNA gene and fungal internal transcribed spacer 1 (ITS1) amplicon sequencing was used to characterize microbial succession during fermentation; the microbial activity was studied using a liquid chromatography-tandem mass spectrometry (LC-MS/MS)-based metaproteomics analysis. Metabolic succession was detected by highperformance liquid chromatography (HPLC) and an ultrahigh-performance liquid chromatography (UPLC)-MS/MS-based metabolomics analysis. Further, we surveyed microbial enzymes involved in the degradation of polysaccharides and the metabolism of phenolic compounds (Fig. 1). This work advanced the production of pu-erh tea and highlights the power of integrated meta-omics approaches in understanding the microbiome in food fermentation ecosystems. method. In total, 36 samples were collected and analyzed, as outlined in Fig. 1 and Table 1. Sensory evaluation revealed that the infusion of raw tea material was slightly astringent in taste and yellow, but after fermentation, the tea infusion became mellow in taste and reddish-brown. This sensory quality change was similar to general fermentation observations of pu-erh tea. It is believed that changes in sensory qualities are closely related to changes in chemical metabolites and caused by microbial activity. Therefore, the composition and activity of microbes in the fermentation process were further investigated by integrated meta-omics approaches.
Microbiota in the fermentation of pu-erh tea. A total of 877,307 bacterial 16S rRNA and 1,740,425 fungal ITS sequences were obtained from 36 samples of tea leaves, and most of the microbial diversity was captured by this analysis (see Fig. S1A and B in the supplemental material). As the fermentation process progresses, bacterial richness and diversity increase, showing a peak in B4 or M6 and decreasing slightly in the last stages of fermentation ( Fig. 2A). Fungal richness and diversity decreased in the initial stage of fermentation, and increased slightly during the middle and final stages of fermentation (Fig. 2B).
Principal-component analysis (PCA) based on the relative abundance (RA) of operational taxonomic units (OTUs) revealed that both bacterial and fungal communities in the raw material (B0 and M0) differed from those of other fermented stages ( Fig. 2C and D); bacterial communities in the final stages of the fermentation of tea leaves (B7, B8, M7, and M8) were similar to each other and yet they differed from those of other stages. However, fungal communities in the final fermentation steps of tea leaves (B8 and M8) were similar to those in other fermented samples (B1, B2, B3, M1, M2, M3, and M4).
Bacterial OTUs were classified into 31 phyla (Data Set S1, sheet 1). At the phylum level, there was a decrease in the proportion of Proteobacteria and an increase in the abundance of Firmicutes. Additionally, the RA of Actinobacteria increased to 15.37% (B4), 24.90% (B8), and 57.51% (M7) (Fig. S1C). More than 80% of bacterial OTUs were assigned to 201 families (Data Set S1, sheet 2). As the fermentation process progressed, proportions of Enterobacteriaceae decreased in FB, whereas they increased in FM. The abundance of Bacillaceae increased at the final stages in both FB and FM with RAs of 44.92% (B8) and 89.05% (M8). Additionally, RAs of Lactobacillaceae were greater than 70% in B5 and B6, and the RA of Pseudomonadaceae was 46.56% in M1 ( Fig. 2E and Fig. S1E). Similarly to our previous findings (20), the RA of Enterobacteriaceae decreased and the proportions of Bacillaceae increased during the fermentation process of pu-erh tea.
Fungal OTUs were classified into four phyla and unclassified fungi. The dominant phylum was Ascomycota, which accounted for 84.76% to 99.94% of the total fungi ( Fig. S1D and Data Set S1, sheet 3). More than 87% of fungal reads were classified into 166 genera; Aspergillus and Debaryomyces were dominant genera in B0 with RAs of  (Fig. 2F, Fig. S1F, and Data Set S1, sheet 4). The dominance of Aspergillus and the presence of Rasamsonia and Thermomyces in the fermentation of pu-erh tea have been previously reported (19,21).
To obtain a measure of microbial association, three OTU cooccurrence networks were constructed. In the bacterial cooccurrence network, most of the OTUs corresponding to Bacillaceae, Comamonadaceae, Pseudoalteromonadaceae, Pseudomonadaceae, Phyllobacteriaceae, and Vibrionaceae cooccurred with others in fermentation. OTUs assigned to unclassified groups in Rickettsiales and Enterobacteriaceae were negatively correlated with other bacteria (Fig. 3A). In the fungal network, most of the OTUs assigned to Thermomyces and Thermoascus have negative correlations with other genera, whereas the OTUs corresponding to Aspergillus, Rasamsonia, Penicillium, and Debaryomyces cooccurred with each other (Fig. 3B). The network of OTUs of 16S RNA genes and ITS sequences showed that bacteria and fungi were mutually exclusive in fermentation. For example, members of the Bacillaceae, Pseudoalteromonadaceae, Comamonadaceae, and Enterobacteriaceae showed negative correlations with fungi in Aspergillus, Rasamsonia, Penicillium, Debaryomyces, and Saccharomycetes (Fig. 3C).
Overview of metaproteomics results. After protein extraction, LC-MS/MS analyses, and a database search, 68 to 1,582 microbial proteins in each repeated analysis and 4,623 and 6,505 unique proteins in FB and FM, respectively, were identified and further annotated, with proteins identified as malate dehydrogenase, superoxide dismutase, and catalase, among others (Data Set S2, sheets 1 and 2). In the Gene Ontology (GO) annotation, the majority of identified proteins categorized as molecular functions were primarily catalytic activity and binding; those categorized as biological process were cellular process and metabolic process; the proteins categorized as cellular components were cell parts and protein-containing complex. The majority of enzyme classes were oxidoreductases, transferases, and hydrolases ( Fig. S2A and Data Set S2, sheets 3 and 4). The major categories identified by the Cluster of Orthologous Group analysis were energy production and conversion, translation, ribosomal structure and biogenesis, posttranslational modification, protein turnover and chaperones, amino acid transport and metabolism, and carbohydrate transport and metabolism ( Fig. S2B and Data Set S2, sheet 5). A total of 116 KEGG pathways were annotated, with the most common pathways identified as glycolysis/gluconeogenesis (ko00010), ribosome (ko03008), oxidative phosphorylation (ko00190), and citrate cycle (ko00020) ( Fig. S2C and Data Set S2, sheet 6). These KEGG pathways grouped into cellular processes, environmental information processing, genetic information processing, and metabolism. KEGG pathways in metabolism were further associated with classes of amino acid metabolism, biosynthesis of other secondary metabolites, and carbohydrate metabolism (Fig. S2C). Overall, the majority of identified proteins were assigned to cellular process and metabolic process in the GO analysis and enriched in pathways belonging to metabolism or genetic information processing. These data support the findings from microbial growth and reproduction.
Metabolic succession in fermentation. The concentrations of 16 characteristic components of tea were measured by HPLC or spectrophotometric methods. We observed three change trends among the results ( (ii) the contents of water extractions (WE), kaempferol, quercetin, myricetin, and (Ϫ)-epicatechin (EC) increased at the initial stage and then decreased significantly after fermentation (P Ͻ 0.05); and (iii) the levels of soluble sugar (SS), gallic acid, and ellagic acid increased significantly after fermentation (P Ͻ 0.05). Additionally, the caffeine content increased significantly (P Ͻ 0.05) in FB but showed no significant change in FM (P Ͼ 0.05).
A total of 11,423 m/z were detected in the metabolomics analysis (Data Set S3, sheets 1 and 2), and the PCA with 67.0% variation showed that the metabolites identified in samples from raw material (B0-1, B0-2, M0-1, and M0-2), the middle stage    (Fig. 4C). After fermentation, the relative levels of 124 and 125 metabolites were decreased significantly in comparisons of B8 and B0 and of M8 and M0, respectively. Among them, relative peak areas of 64 and 57 metabolites, respectively, decreased by more than 10-fold, including EGCG, theaflavin digallate, luteoliflavan, and L-theanine, whereas the relative peak areas of 55 and 52 metabolites significantly increased after fermentation in comparison of B8 and B0 and of M8 and M0, respectively, including margrapine A, kukoamine A, Thr-Trp-OH, Phe-Lys-OH, uridine, ellagic acid, and gallic acid (Data Set S3, sheet 5). HPLC determination verified the increasing contents of gallic acid, which increased 11.63 and 5.94 times in FB and FM, respectively.
Metabolism of phenolic compounds in the fermentation of pu-erh tea. Phenolic compounds possess one or more aromatic rings with one or more hydroxyl groups and generally are categorized as phenolic acids, flavonoids, coumarins, and tannins (27). Phenolic compounds, primarily catechins, are the characteristic chemical component in teas and provide a number of health benefits, such as reducing the incidence of coronary heart disease, diabetes, and cancer (28). Moreover, the oxidation of catechins and the production of oxidation reaction products in fermentation are crucial to the quality of black tea (29). Therefore, understanding the metabolism of phenolic compounds is essential for the investigation of the process and quality control of tea.
In this metabolomics study, 144 phenolic compounds were identified, including catechin 3-O-gallate, gallocatechin, and quercetin (Data Set S3, sheet 7). After fermentation, the relative levels of 73 metabolites decreased, which was in accordance with the decreasing levels of polyphenols and catechins, including EGC, EC, EGCG, C, and ECG, as shown by the spectrophotometric or HPLC analyses. Thus, phenolic compounds were actively metabolized, and the majority of levels of phenolic compounds decreased during fermentation. The decreasing content of phenolic compounds is responsible for the transformation of taste from astringent to mellow.
We hypothesized a metabolic pathway of tea phenolic compounds resulting in the decrease of relative levels of most phenolic compounds and an increase in the content of several compounds including gallic acid, ellagic acid, quercetin, and myricetin in pu-erh fermentation as follows: (i) phenolic glycosides were hydrolyzed or synthesized by GHs and GTs; (ii) gallates were hydrolyzed by tannase and produced gallic acid; (iii) phenolic compounds were oxidized, modified, or degraded by catechol O-methyltransferase, phenol 2-monooxygenase, salicylaldehyde dehydrogenase, salicylate 1-monooxygenase, catechol 2,3-dioxygenases, catechol 1,2-dioxygenase, and quercetin 2,3-dioxygenase. To our knowledge, this is the first report on the enzymes involved in the metabolism of phenolic compounds in pu-erh tea fermentation, which are characteristic compounds in tea and are responsible for the taste and health benefits. Additionally, phenolic compounds are ubiquitously distributed phytochemicals found in most plant tissues and are important for the quality of plant-based foods (31); therefore, the findings in this article may provide interesting insight into other plant-based fermented foods and beverages.
Development of the FFMP. Fermented foods are important societal traditions and are crucial regional products in terms of the economy, as well as being rich in microbiological resources awaiting exploration (32). Understanding the microbiome within fermentation ecosystems is essential for maintaining traditional and artisanal practices in the context of urbanization, designing starter cultures, directing sensory quality, and improving the safety of the consumable products (6). Previously, Parente (33,34) developed FoodMicrobionet, which provides a wealth of information on the structure of food biomes. We suggest developing the Food Fermentation Microbiome Project (FFMP) to study the microbiome within the food fermentation ecosystem using the powerful integration of meta-omics approaches. This work provides an example of a study of microbiomes in a fermented food ecosystem using integrated metabarcoding, metaproteomics, metabolomics, and HPLC approaches.
Conclusion. Microbiomes in two fermentations of pu-erh tea were systematically examined via the integration of metabarcoding, metaproteomics, and metabolomics analyses. We identified the microbial succession and association, microbial activity, and changes in the metabolites during the fermentation of pu-erh tea. We found that microbiota produced CAZymes to degrade plant or fungal polysaccharides for their growth and reproduction, as well as enzymes involved in hydrolysis, oxidization, modification, or degradation of phenolic compounds (Fig. 6). This study advanced our understanding of the fermentation mechanism of pu-erh tea related to the microbial and metabolic succession, as well as the microbial functions during the fermentation of pu-erh tea.

MATERIALS AND METHODS
Fermentation of pu-erh tea and sample collection. Two traditional fermentation processes of pu-erh tea were developed by the Yunnan D Tea Co., Ltd., Yunnan, China, between 10 October and 1 December 2014 ( China, were used as the raw material. The fermentation of pu-erh tea was developed according to the traditional method of spontaneous fermentation; raw materials, water, utensils, and the environment were not sterilized, and no starter was used. Based on the temperature of the tea piles and the experience of the manufacturer, the tea masses were broken down, washed with water to a moisture content of approximately 40%, mixed, and restacked in piles for 3 to 10 days. Samples of tea leaves were collected from the tea piles at five time points before each round of breaking up, mixing, and repiling. Samples of tea leaves were divided into two parts. One part was air dried and subjected to sensory evaluation, according to the protocol described by GB/T 23776-2009 (35), and analysis of the chemical compounds, while the second part was stored at Ϫ80°C. In total, 36 samples were collected and analyzed as outlined in Fig. 1 and Table 1. Detailed approaches are described in Text S1 in the supplemental material.
Metabarcoding of bacterial 16S rRNA gene and fungal ITS sequence. To analyze the taxonomic composition of the bacterial and fungal communities, the universal primer pairs 515F (5=-GTGCCAGCM GCCGCGGTAA-3=) and 907R (5=-CCGTCAATTCMTTTRAGTTT-3=) and ITS1F (5=-CTTGGTCATTTAGAGGAAG TAA-3=) and ITS1R (5=-GCTGCGTTCTTCATCGATGC-3=), which incorporate Illumina adapters and barcode sequences, were used to amplify the V4-V5 hypervariable region of bacterial 16S rRNA genes, as well as the ITS1 of fungal 18S rRNA genes using a two-step amplification procedure. DNA extraction, PCR, and Illumina MiSeq sequencing (2-by 150-bp reads) were performed by TinyGene Technology Co., Ltd. (Shanghai, China). Each sample was extracted for two replicates, and each extraction was analyzed twice. Analysis of OTU cooccurrence networks was developed using the CoNet application (36) on Cytoscape 3.7.1 (37). The detailed approaches are described in Text S1 in the supplemental material.
Metaproteomics experiments. The microbial proteins in tea leaves were extracted by Tris-HClphenol and methanol precipitation, measured using the Bradford method with bovine serum albumin as a standard, and validated with sodium dodecyl sulfate-polyacrylamide electrophoresis (as described in our previous report [19]). For each sample of tea leaves, three independent extractions were carried out. A total of 200 g of protein was digested with trypsin according to the filter-aided sample preparation protocol (38). Liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis of each replicate of peptide extracts was performed using an Easy-nLC1000 coupled to a QExactivePlus mass spectrometer (Thermo Fisher Scientific, Bremen, Germany). Raw data were processed using Thermo Proteome Discoverer software version 1.4 (Thermo Fisher Scientific, Bremen, Germany) with the default settings. The MS/MS data were queried against the UniProt database (http://www.uniprot.org/) with the following search parameters: carbamidomethylation of cysteine as the fixed modification, oxidation of methionine and deamidation of glutamine and asparagine as variable modifications, a maximum of two missed cleavages, a precursor ion mass tolerance of 10 ppm, and an MS/MS tolerance of 0.05 Da. Decoy database searches were performed with a false-discovery rate (FDR) cutoff of 1%. GO annotations for the identified proteins were assigned according to those reported in the UniProt database. COG annotations of identified proteins were computed using eggNOG-Mapper based on eggNOG 4.5 orthology data (39,40). The CAZymes annotation was developed by dbCAN (41). The KEGG pathway was annotated using the KEGG (Kyoto Encyclopedia of Genes and Genomes) Automated Annotation Server (KAAS) using the bidirectional best hit BLAST method (https://www.genome.jp/tools/kaas/) (42). Detailed approaches are described in Text S1 in the supplemental material.
Data availability. The sequencing data of bacterial 16S rRNA genes and the fungal ITS1 of 18S rRNA gene are available at the Sequence Read Archive under project code SRP139059. The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium (http://proteomecentral .proteomexchange.org) via the iProX partner repository (43) with the identifier PXD012223.