Microbial Diversity Analysis of Fermented Mung Beans (Lu-Doh-Huang) by Using Pyrosequencing and Culture Methods

In Taiwanese alternative medicine Lu-doh-huang (also called Pracparatum mungo), mung beans are mixed with various herbal medicines and undergo a 4-stage process of anaerobic fermentation. Here we used high-throughput sequencing of the 16S rRNA gene to profile the bacterial community structure of Lu-doh-huang samples. Pyrosequencing of samples obtained at 7 points during fermentation revealed 9 phyla, 264 genera, and 586 species of bacteria. While mung beans were inside bamboo sections (stages 1 and 2 of the fermentation process), family Lactobacillaceae and genus Lactobacillus emerged in highest abundance; Lactobacillus plantarum was broadly distributed among these samples. During stage 3, the bacterial distribution shifted to family Porphyromonadaceae, and Butyricimonas virosa became the predominant microbial component. Thereafter, bacterial counts decreased dramatically, and organisms were too few to be detected during stage 4. In addition, the microbial compositions of the liquids used for soaking bamboo sections were dramatically different: Exiguobacterium mexicanum predominated in the fermented soybean solution whereas B. virosa was predominant in running spring water. Furthermore, our results from pyrosequencing paralleled those we obtained by using the traditional culture method, which targets lactic acid bacteria. In conclusion, the microbial communities during Lu-doh-huang fermentation were markedly diverse, and pyrosequencing revealed a complete picture of the microbial consortium.

The manufacturing process of Lu-doh-huang is divided into 4 stages, and each stage lasts for 4 months ( Fig. 1 and 2). Stage 1 runs from about the beginning of September to the end of the following January. During this stage, mung beans are mixed with various Chinese herbal medicines, including powdered Ganoderma lucidum, Taiwanofungus camphoratus, Coptis chinensis Franch., Houttuynia cordata Thunb., and Glycyrrhiza uralensis Fisch. Small-diameter holes are drilled into sections of live bamboo. The mixture of mung beans then is stuffed through these holes in the bamboo by using a funnel; once the section is filled, these holes are corked and sealed tightly with tape. During stage 2, the bean-filled bamboo sections are sawn open and soaked in a spontaneously fermented soybean solution; these bamboo sections are transferred to a pool of slowly running spring water to soak during stage 3. At stage 4, the bamboo sections are cleaved into two so that the mung beans can be removed, rinsed in spring water, and sun-dried during the day. At night, the mung beans are soaked in an aqueous mixture of Chinese herbal medicines or put outside to expose them to the dew. These steps are repeated several times until the mung beans are dried completely to become the final product. Although a wide variety of fermented soybean products are used worldwideincluding chungkokjang (Korea), kinema (India), miso (Japan), natto (Japan), shoyu (China and Japan), and tempeh (Indonesia)none results from a fermentation process similar to that for Ludoh-huang. Whereas the medical applications of Lu-doh-huang have been explored in several studies, the composition of the bacterial communities during its manufacture remains unknown. In the current study, we used pyrosequencing to assess the microbial diversity of Lu-doh-huang during its production by fermentation and applied the traditional culture method to identify the predominant lactic acid bacteria (LAB) during Lu-doh-huang fermentation.

Sample Collection and pH Analysis
Throughout the 4-stage process, we collected samples of the mung beans inside bamboo sections and the liquids in which the bamboo sections were soaked. At least 3 samples each of beans and liquid were collected at each of 7 points during the manufacturing process; like samples were pooled for further analysis. During stage 1, mung beans were fermented in sections of live bamboo; samples were collected after 2 weeks (S1-2W), 1 month (S1-1M), and 4 months (S1-4M) of fermentation. During stage 2, the bamboo sections containing mung beans were soaked in fermented soybean liquid (FSL) for 4 months (the S2-4M sample); during stage 3, the bamboo sections were soaked in slowly running spring water (RSW) for 4 months (the S3-4M sample). During stage 4, the mung beans were removed from the bamboo sections and then repeatedly sun-dried during the day and exposed to dew (or Chinese herbal medicine) during the night for 4 months (S4-4M). The fermented mung beans then were dried to yield the final product, Lu-doh-huang (LDH). Samples of the unfermented material (UFM) and the FSL and RSW in which the filled bamboo sections were soaked were collected also. The samples were stored on ice during transport to the laboratory for analysis within 7 days of sampling. The pH of the samples was measured (model PHB-9901, DO Microelectronic pH-Vision meter, Ai-On Industrial Corporation, Taipei, Taiwan).

PCR Amplification for Pyrosequencing
The V1-V2 hypervariable regions of the 16S rRNA genes of bacteria present at each sampling point were amplified individually by using nested PCR as previously described [6]. For the first round of amplification, primers 27f (59-AGRGTTTGA-TYMTGGCTCAG-39) and 338r (59-TGCTGCCTCCCGTAG-GAGT-39) were used. Each reaction (volume, 25 mL) contained 10 ng of extracted DNA, 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl 2 , 200 mM deoxynucleoside triphosphate (dNTP) mixture, 5 pmol of each primer, and 0.625 U ExTaq HS (Takara Bio, Shiga, Japan). Thermocycling conditions were: 1 cycle of 98uC for 2.5 min; 15 cycles of 98uC for 15 s, 50uC for 30 s, and 72uC for 20 s; and a final cycle of 72uC for 5 min. The second round of PCR was performed by using the primers Pyro-27f (59-CCATCTCATCCCTGCGTGTCTCCGACtcagN 10 AGRGTTT-GATYMTGGCTCAG-39) and Pyro-338r (59-CCTATCCCCTGTGTGCCTTGGCAGTCt-cagTGCTGCCTCCCGTAGGAGT-39). The underlined sequences are dictated by the requirements of the 454 Sequencing System (Roche Diagnostics, Branford, CT, USA); the nucleotides in lowercase letters are sequence keys used for amplicon sequencing; italicized sequences are those for universal primers for bacteria; and N 10 designates the 10-base multiplex identifiers (Roche Diagnostics) used to label each PCR product. The second round of PCR amplification was performed in a 50-mL volume containing 0.5 mL of the product of the first-round of PCR, 10 pmol of each primer, and 1.25 U Ex Taq HS (TaKaRa Bio., Shiga, Japan); all other components were the same as for the first round of PCR. Cycling conditions for the second round of PCR were: 1 cycle of 98uC for 2.5 min; 15 cycles at 98uC for 15 s, 54uC for 30 s, and 72uC for 20 s; and a final cycle of 72uC for 5 min.
The quantity and quality of amplicons were determined by using the Quant-iT PicoGreen dsDNA Assay Kit (Molecular Probes, OR. USA) and Agilent DNA 1000 chip, respectively, on a BioAnalyzer 2100 (Agilent Technologies, CA, USA). Each amplicon was purified separately by using Agencourt AMPure XP beads (Beckman Coulter, Brea, CA, USA) according to the manufacturer's protocol. Purified amplicons were quantified, diluted to 1610 9 molecules/mL, and underwent a second purification step by using Agencourt AMPure XP beads (Beckman Coulter). Samples with different sample-specific barcode sequences were quantified and diluted to 1610 8 molecules/mL each and pooled proportionally. The quality of the amplicon pool was assessed by using a High-Sensitivity DNA chip (Agilent Technologies) and subsequently subjected to emulsion PCR by using a GS emPCR Lib-L Kit (Roche Diagnostics). Pyrosequencing was performed on a GS Junior System (Roche Diagnostics) by using Titanium chemistry (Roche Diagnostics) according to the manufacturer's instructions.

Analysis of Pyrosequencing Data
Raw Fasta data were trimmed by using FASTX-Toolkit [7] (Q.25); reads were recognized as derived from specific samples by their unique multiplex identifier. After removing the multiplex identifier and primer sequences for each read, only reads longer than 310 nucleotides were chosen for taxonomic assignment by using BLASTn [8]. High-quality trimmed reads were used as BLASTn queries against the 9701 full-length ($1200 nucleotides) high-quality 16S rRNA sequences from isolated typed strains (September 2012, RDP release 10, update 30) [9]. In BLASTn results, only the best hit queried by each read was considered. Only reads with more than 97% sequence identity were assigned to the corresponding species. Aligned sequences were clustered into operational taxonomic units (OTU) defined at 3% sequence distance by using the complete-linkage clustering tool in RDP [9]. The Shannon-Weaver index [10], Chao1 index, and evenness and rarefaction analyses also were completed by using the web-based services of RDP.
The relative abundance of sequence reads at the bacterial genus level for seven samples were studied by principal component analysis (PCA). Hierarchical clustering was performed according to the Euclidean distance and complete linkage method. The data used for the clustering analysis were the relative abundance matrix, in which each row corresponds to a genus, and each column corresponds to a sample. Before clustering, data were filtered to eliminate genera with the sum of relative abundance across seven samples lower than 1%; remaining data were log 10-transformed. A heatmap of the results of the clustering analysis and PCA were generated by using R/Bioconductor (www. R-project.org; www.bioconductor.org).

Isolation and Enumeration of LAB
Mung beans collected from the various processing stages were washed with 9-fold volume of phosphate buffered saline. After vigorous vortexing, serial 10-fold dilutions were performed, and 0.1-mL aliquots of the appropriate dilutions were plated onto acidic MRS agar plates (pH 5.4; Diagnostic Systems, MD, USA) supplemented with 0.001% sodium azide and cycloheximide for LAB analysis. The plates were incubated at 30uC for 3 days in an anaerobic chamber (Coy Laboratory Products, Grass Lake, MI, USA; atmosphere comprising 85:5:10 N 2 :CO 2 :H 2 ). Colonies with distinct morphologies (that is, color, shape, and size) were selected and purified by streaking at least twice on MRS (pH 6.8) agar plates. The isolates were stored at -80uC in Nutrient Broth (Diagnostic Systems) containing 10% dimethyl sulfoxide.

DNA Extraction
LAB isolates were cultured in MRS broth (pH 6.8) at 30uC for 1 to 2 days in an anaerobic chamber (Coy Laboratory Products) and harvested by centrifugation at 14,0006g for 3 min. For the genomic DNA extraction, the solid samples collected at the different processing stages were ground into powder, and diluted with 9-fold sterile saline (0.85% NaCl). After vigorous vortexing, 0.2-ml aliquots were used, as well as the liquid samples. The genomic DNA of LAB isolates or the total DNA of these samples was extracted as previously described [11]. Extracted DNA was dissolved in TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) for further analyses. To obtain total DNA, mung beans were suspended with 9-fold volume of phosphate buffered saline; total DNA was extracted from 200-mL aliquots of the suspension.

RAPD Fingerprinting
RAPD-PCR analysis using random primers A (59-CCGCAGC-CAA-39), B (59-AACGCGCAAC-39), and C (59-GCGGAAA-TAG-39) was performed as previously described [11]. The PCR products were separated by electrophoresis on a 1% agarose gel. Isolates with the same pattern were grouped, and a single representative strain from each group was chosen for subsequent gene analyses.

Sequencing and Phylogenetic Analyses of 16S rRNA, pheS, and recA Genes
Approximately 1500 bases of each 16S rRNA gene sequence was amplified as previously described [11]. The primer sets and reaction conditions used for pheS gene amplification were those previously described [12,13]. To identify strains in the Lactobacillus plantarum group (including L. paraplantarum, L. pentosus, and L. plantarum), recA sequences were used [14]. The PCR products were sequenced at the National Yang-Ming University Genome Research Centre. The methods for sequence assembly, similarity searches, and phylogenetic analyses have been described previously [15].

Pyrosequencing-based 16S rRNA Profiling of Lu-dohhuang
High-throughput DNA analysis was applied to analyze changes in the total microbial composition during the manufacture of Ludoh-huang. PCR yields from the UFM, S4-4M, and LDH samples were insufficient for further analysis due to the submarginal DNA-yield. A total of 71,792 sequence reads of the 16S rRNA gene were obtained from a single run by using barcoded pyrosequencing. We obtained a total of 71,623 highquality reads, which had an average length of 340 bases and yielded approximately 24 Mb of sequence information. After barcode sorting, the read number of each sample varied from 6,622 to 14,473, with an average of 10,232 (Table 1). According to the Shannon-Weaver index, microbial diversity decreased as fermentation progressed during stage 1. However, the indices from S2-4M and S3-4M, corresponding to stages 2 and 3, respectively, were higher than those from stage 1. The microbial diversities of FSL and RSW were both higher than those of the samples that soaked in these liquids. The species evenness and rarefaction curves supported this finding (Fig. S1). We used rarefaction curves of reads at the 97% similarity level to estimate the richness of the bacterial communities. Except for the FSL and RSW samples, the OTUs detected were sufficient to represent the microbial composition present. In FSL, the numbers of genera and species detected were as many as 162 and 291, whereas RSW yielded 116 genera and 189 species. However, the sequence reads of these two samples were not sufficient to describe the entire microbial community, given that neither of the rarefaction curves reached a plateau.
For all genera detected, 48 genera occupied 84.8% to 99.2% of the total bacterial community which the minimum incidence of 1% in a genus in each sample. To analyze the bacterial community associated with Lu-doh-huang fermentation in detail, we performed hierarchical clustering by using the relative abundance of genera (Fig. 4). The microflora of the environmental samples, FSL and RSW, formed a cluster separate from those representing the bacterial communities in mung bean samples obtained from bamboo sections (that is, S1-2W, S1-1M, S1-4M, S2-4M, and S3-4M). The dominant OTUs inside the bamboo sections were clearly dissimilar to those of environmental samples. Samples S1-1M, S1-4M, and S2-4M were close to each other in the hierarchical clustering, and the genus Lactobacillus was the strongest influencing factor for clustering. The remaining samples (S1-2W, FSL, RSW, and S3-4M) were well separated from other clusters. In addition, we performed PCA to characterize the microbial community succession of Lu-doh-huang (Fig. 5). A projection of the first two principal components explained about 75% (61.3% and 13.6%) of the total variance, with strong separation by region.

pH and Cell Counts of LAB of Lu-doh-huang Samples
The pH of the unfermented starting material (UFM), which contained mung beans and various Chinese medicines, was 5.4 (Table 3). After this mixture was stuffed into bamboo sections for 2 weeks (S1-2W), the pH increased slightly to 5.9. However, pH started to decrease gradually and 4 months later, it was 3.9 (S1-4W). During stages 2 and 3 (S2-4M and S3-4M), pH increased to approximately 5.0. The pH of the FSL and RSW samples was 5.5 and 6.9, respectively.
The traditional culture method detected no LAB in the UFM, S4-4M, and LDH samples (Table 3). After anaerobic fermentation for 2 weeks inside bamboo sections, the cell counts of LAB increased to the level of 10 8 cfu/g. This cell count remained stable throughout the manufacturing process, until the fermented mung beans were removed from the bamboo sections. However, sample S4-4M lacked LAB, which were isolated from FSL and RSW at a level of about 10 4 cfu/mL.

Identification and Distribution of LAB
LAB strains were then isolated from each sample according to their colony morphologies and analyzed by RAPD-PCR for grouping and 16S rRNA gene sequencing for identification. In total, 218 strains were isolated (7 from UFM, 35 from S1-2W, 16 from S1-1M, 24 from S1-4M, 16 from FSL, 72 from S2-4M, 16 from RSW, and 32 from S3-4M). No colonies grew on MSR agar streaked with the LDH sample. On the basis of the resulting RAPD profiles, 218 isolates were categorized into 82 different groups (5 from UFM, 10 from S1-2W, 9 from S1-1M, 4 from S1-4M, 9 from FSL, 25 from S2-4M, 9 from RSW, and 11 from S3-4M). A representative strain was chosen from each of the 82 groups for partial sequencing of the 16S rRNA gene (500 nucleotides containing the V1 to V3 regions) [16]; the sequences obtained were identified as belonging to 4 genera (Enterococcus, Lactobacillus, Lactococcus, and Pediococcus) and 24 species. Most of the isolates were Lactobacillus, comprising 19 different species. The strains from UFM were not LAB. Sample S1-2W contained 6 different LAB species; Enterococcus casseliflavus and Lactococcus lactis were the dominant species, with cell counts as high as 10 8 cfu/g ( Table 4). As the fermentation proceeded, the number of LAB species decreased, and the main species shifted to Lactobacillus brevis and L. plantarum (S1-1M and S1-4M). After the bamboo sections had been opened to soak in the fermented soybean liquid for 4 months (S2-4M), the diversity of LAB had changed dramatically. The 10 LAB species obtained were present in similar numbers, and L. plantarum was the only species that remained from stage 1. None of the LAB species isolated from FSL was present in S2-4M. Similarly, except for L. brevis, none of the LAB in RSW was found in S3-4M. LAB cell counts increased slightly between stages 2 and 3, and L. brevis and L. pentosus predominated during both stages.
Among the Lactobacillus species identified, 6 representative strains isolated from S1-2W and S3-4W could not be identified conclusively by using the full-length 16S rRNA gene sequences because these sequences were more than 99% or less than 97% similar to those of known species in the databases. Therefore, we compared pheS gene sequences of these 6 strains. Because the pheS sequence identities of these strains were all lower than 90% with recognized species (data not shown), these isolates were tentatively identified as Lactobacillus sp. A, B, C, and D.

Discussion
In this study, we used pyrosequencing of tagged 16S rRNA gene amplicons and the traditional culture method to reveal changes in microbioflora of Lu-doh-huang during its production. Lu-dohhuang is a unique regional and alternative medicine that is especially popular in southern Taiwan. To produce Lu-dohhuang, mung beans are mixed with other Chinese herbal medicines, fermented anaerobically, soaked in liquid herbal medicines repeatedly, and then dried completely. A similar product is Semen Sojae Praeparatum in mainland China. The processing of Lu-doh-huang and Semen Sojae Praeparatum is comparable: both combine a type of bean (mung beans and black beans, respectively) with other herbal medicines and undergo natural fermentation. A key difference between these products is the type of fermentation used in their manufacture, that is, Ludoh-huang undergoes strictly anaerobic fermentation (inside bamboo sections), whereas the production of Semen Sojae Praeparatum involves aerobic fermentation. However, the microflora of Semen Sojae Praeparatum has not been analyzed extensively [17].
Natural fermentation is a common production method for Chinese herbal medicines that has been used for almost 2,000 years. Many medicines produced through natural fermentation are still popular, including Massa Medicata Fermentata, medicated leaven, Semen Sojae Praeparatum, and Chinese gall leaven [18]. Microbial enzymes present during fermentation are thought to convert or detoxify bioactive compounds in the medicinal herbs to enhance the health benefits of the final product [18,19]. For example, gastrointestinal bacteria transform bioactive ingredients of herbal medicines so that they become easily absorbed or    . Cell counts are given as Log10 cfu/g or cfu/mL, depending on the sample (mung beans or soaking liquid). doi:10.1371/journal.pone.0063816.t004