Tracing microbiota changes in yamahai-moto, the traditional Japanese sake starter

Abstract

Sake is made from steamed rice, malted rice, and water. Sake production begins with the preparation of a small-scale starter (moto); the quality of moto significantly influences the flavor and richness of sake. In the traditional starter, yamahai-moto, the growth of naturally occurring lactic acid bacteria represses the putrefactive micro-organisms, whereas in the modern starter, sokujo-moto, this is achieved by adding lactic acid. In this study, the successive change in bacterial flora of yamahai-moto was analyzed by pyrosequencing 16S ribosomal RNA genes. Lactobacillus was dominant throughout the process (93–98%). Nitrate-reducing bacteria that have been generally assumed to be the first colonizers of yamahai-moto were scarcely found in the early stage, but Lactobacillus acidipiscis dominated. Lactobacillus sakei drastically increased in the middle stage. This is the first report, though one case study, to show how the early stage microbiota in Japanese yamahai-moto is varyingly controlled without nitrate-reducing bacteria using next-generation sequencing.

Graphical abstract

Microbiota change during the fermentation of traditional Japanese sake starter yamahai-moto.

Key words:

• 16S rDNA pyrosequencing
• fermented food microbiota
• lactic acid bacteria
• Japanese sake starter
• yamahai-moto

Japanese sake production is conducted by fermenting steamed rice and malted rice (kome-koji) with water.^1–4) Kome-koji is a solid-fermented steamed rice prepared using Aspergillys oryzae as a saccharifier to degrade rice starch in the special brewery’s room (koji-muro). Kome-koji is then used for preparing a starter brew, called moto, by manually mixing it with steamed rice and water in an open tank. Modern Japanese sake breweries generally use a type of starter called sokujo-moto, which includes lactic acid to maintain a low pH, thus preventing microbial contamination. Brewer’s yeast Saccharomyces cerevisiae is immediately added, to stimulate alcohol formation.¹⁾ Sokujo-moto preparation requires approximately half a month to ensure sufficient viability of yeast to stably ferment the main sake brew (Fig. 1). Alternatively, a traditional starter, yamahai-moto (also called kimoto), is made over a longer period of approximately 1 month. In this starter, lactic acid is formed through the growth of lactic acid bacteria (Fig. 1). The preparation of yamahai-moto therefore starts at neutral pH, and there is a higher risk of putrefaction by an unexpected growth of spoilage micro-organisms. However, yamahai-moto is still produced by some Japanese breweries, as the sake made using this type of starter has a specific flavor, owing to the complex microbiota succession during the moto preparation.

Fig. 1. Production scheme and generally regarded microbial succession for (A) yamahai-moto and (B) sokujo-moto.

Notes: Steamed rice and kome-koji are mixed in water, then incubated for approximately 1 month for yamahai-moto and half a month for sokujo-moto. In sokujo-moto, lactic acid and brewer’s yeast are added at the starting point, whereas yamahai-moto requires the indicated microbial succession before yeast addition. Microbial succession model of yamahai-moto adapted from Akiyama.⁴³

It is known that the early fermentation stage of yamahai-moto contains wild yeasts, film-forming yeasts, and nitrate-reducing bacteria derived from the starting materials and water (Fig. 1). Nitrate-reducing bacteria convert nitrate dissolved in the water into nitrite, which suppresses the growth of contaminating micro-organisms. Afterward, lactic acid cocci (Leuconostoc) and bacilli (Lactobacillus) grow, reducing the medium’s pH by lactic acid formation, and suppressing the growth of the above first colonizers. This succession occurs until 10−15 days after the start of moto preparation; at this stage the yeast culture is generally added to produce alcohol. Numerous reports have demonstrated, by culture-dependent microbiota analysis, that Lactobacillus sakei is the dominant species in kimoto and yamahai-moto.^{2, 5–10)} Masuda et al.¹¹⁾ revealed the presence of L. sakei and Leuconostoc mesenteroides in yamahai-moto through denaturing gradient gel electrophoresis of polymerase chain reaction (PCR)-amplified 16S ribosomal RNA gene (rDNA). However, quite recently, Bokulich et al.¹²⁾ demonstrated, using next-generation 16S rDNA sequencing, that several Lactobacillus species, including Lactobacillus plantarum, were contained in kimoto but that L. sakei was absent in their analyses. To reveal the composition of yamahai-moto microbiota in more detail, culture-independent next-generation sequencing analyses with broad data coverage should be rigorously undertaken.^13–17)

We analyzed a sample of yamahai-moto bacterial flora by 16S rDNA-based pyrosequencing. Our data established the dominant bacteria as L. sakei, but, unexpectedly, we could not confirm the presence of nitrate-reducing bacteria, such as Pseudomonas, Achromobacter or Enterobacter, at the early fermentation stage. Based on the results of this one case study and the previous studies, we discuss the possibility of microbial communication in the yamahai-moto fermentation medium.

Materials and methods

Samples

Yamahai-moto was prepared by Shata-shuzo Co. Ltd. (Ishikawa, Japan). Kome-koji and steamed rice were mixed in water taken from a clean well and incubated. The moto was prepared over a four-week period, during which yeast (S. cerevisiae) was added at 13 day. Samples were collected 0–17 days after the start of fermentation for viable counts, acid degree, Baumé degree, pH, and lactic acid concentration; samples taken at 1, 3, 6, 11, and 17 days were used for pyrosequencing analysis. The Baumé degree was measured using a commercially available standard Baumé scale hydrometer, and the acid degree was expressed as the volume (in mL) of 0.1 M sodium hydroxide solution required to neutralize 10 mL of filtered yamahai-moto extract. Lactic acid concentration was measured using a RQflex® plus 10 system with Reflectoquant^® lactic acid test strips (Merck).

Pyrosequencing

The variable regions V1 and V2 were used to pyrosequence bacterial 16S rDNA. As reported previously, these regions could be used to classify lactic acid bacteria genera and species.^18–22) The primers used for PCR amplification were F8 (5′-AGAGTTTGATCATGGCTCAG-3′, forward) and R357 (5′-CTGCTGCCTYCCGTA-3′, reverse), whose 5′-ends were flanked by specific adaptors (5′-CGTATCGCCTCCCTCGCGCCATCAG-3′ forward, and 5′-CTATGCGCCTTGCCAGCCCGCTCAG-3′ reverse) that are applicable to the pyrosequencing equipment used, namely, the Genome Sequencer FLX® system (Roche, Basel, Switzerland). Short tag sequences (5′-CACGT-3′, 5′-CATAG-3′, 5′-CGATA-3′, 5′-CTATG-3′, and 5′-TACGA-3′) were inserted between the adaptor and primer sequences. These tags allowed reads from different fermentation periods to be sorted in a single run (1, 3, 6, 11, and 17 days after the start of fermentation, respectively). To prepare the template DNA, the samples were suspended in sterilized 0.85% (w/v) sodium chloride solution, followed by vigorous shaking. The resulting suspensions (equivalent to 80 mg yamahai-moto) were used for DNA extraction using EZ-Extract® for DNA (AMR Inc., Gifu, Japan) and a FastPure® DNA kit (Takara Bio, Otsu, Japan). The PCR experiments were performed using PrimeSTAR® GXL DNA polymerase (Takara Bio) with 18–25 cycles of denaturation (94 °C, 30 s), annealing (55 °C, 30 s), and extension (68 °C, 1 min). The average read length (368, 368, 375, 375, and 374 bp for 1, 3, 6, 11, and 17 days, respectively) sufficiently covered the targeted variable regions.

Analysis of pyrosequence reads

Sequence reads were analyzed using the Ribosomal Database Project (RDP) Classifier.²³⁾ Reads obtained in the FASTA format were assigned to genus levels with an 80% confidence threshold.²⁴⁾ Chimeric reads were excluded by analyzing the sequences using the DECIPHER server.^25,26) Species were determined using the Basic Local Alignment Search Tool (BLAST). Assignment of operational taxonomic units (OTUs), and calculation of the Shannon–Wiener index (taxonomic diversity) were carried out using the FastGroupII© program.²⁷⁾ To determine OTUs, an identity threshold of 97% was set and reads shorter than 300 bp were excluded. Rarefaction curves were drawn using R version 3.1.0 with the vegan package.^28,29)

Real-time quantitative PCR

The bacterial population in each sample was determined by real-time quantitative PCR (qPCR). The DNA samples, prepared for pyrosequencing as described, were reused as qPCR templates. The 16S rDNA V3–V4 regions were amplified using 331-f (5′-TCCTACGGGAGGCAGCAGT-3′) and 797-r (5′-GGACTACCAGGGTATCTAATCCTGTT-3′) as forward and reverse primers, respectively.^30–32) 5′-FAM-CGTATTACCGCGGCTGCTGGCAC-TAMRA-3′ was used as a TaqMan® probe. Reaction mixtures were prepared using TaqMan® Gene Expression Master Mix (Applied Biosystems, Foster City, CA, USA), and run on a 7300® real-time PCR system (Applied Biosystems), according to the manufacturer’s instructions.

16S rDNA sequence analyses of isolated strains

Bacterial strains were isolated from yamahai-moto at each fermentation stage (0–17 days after the start of fermentation). The strains were cultured on de Man, Rogosa and Sharpe (MRS) agar plates (Becton, Dickinson and Co., Franklin Lakes, NJ, USA), and viable counts were recorded (expressed as colony forming units, cfu, per gram of yamahai-moto). Genomic DNA was extracted from randomly selected isolates using a Wizard® genomic DNA extraction kit (Promega, Fitchburg, WI, USA). Whole regions of 16S rDNA were amplified using the oligonucleotides F8 and R1510 (5′-ACGGYTACCTTTGTTACGACTT-3′) as the primer pair. The amplified fragments were purified using a QIAquick® PCR purification kit (Qiagen, Venlo, Netherlands) and directly sequenced.

Nucleotide sequence accession number

Pyrosequencing reads were deposited in the Sequence Read Archive of the DNA Data Bank of Japan (DDBJ) under accession number DRA002607.

Results

Fermentation during yamahai-moto preparation

We firstly observed the Baumé and acid degrees and viable counts detected in the time-course samples of yamahai-moto (Fig. 2(A)). Acid degree significantly increased around 5–8 days after the start of fermentation, with accompanying increase of viable counts on Lactobacillales-selective MRS medium, indicating that lactic acid bacteria contributed to the acid formation. A significant increase in lactic acid concentration (1600 mg/L) and decrease in pH (4.48), observed at 6 day, confirmed this (Fig. 2(B)). The Baumé degree slightly increased during maturation (until 12 day) but drastically decreased after 17 day, owing to a marked formation of ethanol by the brewer’s yeast added at 13 day (Fig. 2(A)). Vigorous alcohol formation acted as a bactericide to reduce viable counts to almost zero around 16–18 days; similar succession has been well known in the yamahai-moto fermentation.⁹⁾ The continuous increase in acid degree after 18 day could therefore be attributed to the action of yeast, not to the previously observed bacteria.

Fig. 2. Fermentation properties of yamahai-moto.

Notes: Viable counts on MRS agar plate (closed circles), acid degree (open triangles), and Baumé degree (open squares) are shown in (A), and lactic acid concentration (closed triangles) and pH (open circles) are shown in (B).

Yamahai-moto microbiota revealed by 16S rDNA pyrosequencing analysis

The samples taken at 1, 3, 6, 11, and 17 day were analyzed for microbiota, as determined by pyrosequencing 16S rDNA. The taxonomic richness and diversity indices are summarized in Table 1. A marked decrease in species richness (number of OTUs) and diversity (Shannon–Wiener index) were observed at 6 day (OTUs, from 1033 to 668; Shannon–Wiener, from 2.30 to 1.48). These indicated that the drastic taxonomic change took place at this time point.

Table 1. Taxonomic richness and diversity of yamahai-moto samples.

Fermentation time (d)	Reads^a	Number of OTUs	Shannon-Wiener index
1	21,560	963	2.19
3	23,436	1033	2.30
6	22,685	668	1.48
11	23,754	738	1.69
17	18,004	725	1.96

^a Reads shorter than 300 bp were excluded.

As shown in Table 2, a broad variety of genera (44 in total) was observed. Rarefaction analysis based on the richness of the genera showed sufficiently saturating curves as shown in Fig. 3, indicating that enough reads number was sampled from the pyrosequencing data-sets to assess total bacterial community. Throughout the fermentation period, the major constituent was Lactobacillus, which accounted for more than 90% (93–98%, Table 2 and Fig. 4(B)) of the total population. Species-level BLAST searches revealed that Lactobacillus acidipiscis significantly dominated the microbiota at 1 and 3 days (46–50% of total reads, Fig. 4(C)), but that the major species afterward was L. sakei (74–91% of total reads, 6, 11, and 17 days, Fig. 4(C)). A Lactobacillus sp. closely related to L. acidipiscis or pobuzihii formed a secondary significant component at the early stage (23–25%, 1 and 3 days), and Lactobacillus plantarum/pentosus/paraplantarum simultaneously were present in lower proportions (11–12%). Interestingly, a small amount of salt-tolerant lactic acid bacteria Tetragenococcus was detectable in the early stage (4–5%, 1 and 3 days, Fig. 4(B)), although this yamahai-moto did not include salt. At 6 day, Tetragenococcus decreased and Leuconostoc increased (representing 6% of total population). Staphylococcus was also detected in small amounts in the early stage (2%, 3 day). Notably, nitrate-reducing bacteria known to be included in early stage microbiota (e.g. Pseudomonas, Achromobacter, and Enterobacter, Table 2) were scarcely identified (only 1–7 reads of Pseudomonas were detected in each sample).²⁾

Table 2. Sequence read statistics of yamahai-moto samples and their phylogenetic classification, by RDP classifier.

Data-sets	Fermentation time (d)					Total
	1	3	6	11	17
(I) Untrimmed reads	21,613	23,465	22,682	23,767	17,999	109,526
(II) Reads excluded by bootstrap threshold of 80%^a	2,684	2,573	155	350	545	6,307
(III) Reads other than bacteria^b	216	265	26	47	120	674
(IV) Trimmed reads [(I)-(II)-(III)]	18,713	20,627	22,501	23,370	17,334	102,545
Acidovorax	1	1	0	0	0	2
Acinetobacter	25	33	3	2	18	81
Arsenicicoccus	1	0	0	0	1	2
Arthrobacter	3	3	0	0	0	6
Brachybacterium	2	2	0	0	2	6
Brevundimonas	0	1	0	0	0	1
Caulobacter	1	0	0	0	0	1
Chryseobacterium	7	4	0	1	2	14
Comamonas	1	1	0	0	0	2
Corynebacterium	0	1	0	0	0	1
Curtobacterium	0	0	0	0	1	1
Deinococcus	0	1	0	0	0	1
Dermacoccus	3	1	0	0	0	4
Dietzia	3	0	0	0	1	4
Enhydrobacter	1	0	1	0	1	3
Enterococcus	0	0	0	0	1	1
Exiguobacterium	4	2	3	7	3	19
Halomonas	0	2	0	0	0	2
Herbaspirillum	23	41	2	5	7	78
Kocuria	0	0	0	0	1	1
Lactobacillus	17,543	19,132	21,122	22,848	16,792	97,437
Lactococcus	1	1	0	1	0	3
Leuconostoc	2	6	1,254	253	213	1,728
Massilia	1	0	0	0	0	1
Methylobacterium	14	3	0	0	1	18
Microbacterium	15	4	0	0	3	22
Novosphingobium	0	1	0	0	0	1
Pantoea	0	0	4	3	1	8
Paracoccus	2	3	0	0	0	5
Pediococcus	0	1	0	0	0	1
Peptoniphilus	0	1	0	0	0	1
Prevotella	0	1	0	0	0	1
Propionibacterium	2	3	0	0	3	8
Propionimicrobium	0	1	0	0	0	1
Pseudomonas	3	7	1	2	1	14
Raoultella	1	2	0	0	0	3
Rothia	1	1	0	0	0	2
Sorangium	1	0	0	0	0	1
Sphingomonas	7	4	0	0	9	20
Staphylococcus	141	442	63	131	82	859
Tepidimonas	6	0	0	0	0	6
Tetragenococcus	897	920	47	114	191	2,169
Variovorax	1	0	0	0	0	1
Yersinia	0	1	1	3	0	5

^a Sequence reads with bootstrap value less than 80% were not assigned to a bacterial genus.

^b Archaea and chloroplasts.

Fig. 3. Rarefaction curves for each yamahai-moto pyrosequencing data-set.

Notes: Rarefaction analysis was carried out for respective data-sets of yamahai-moto time-course samples. The vertical axis shows the number of genera detected after sampling the number of sequence reads shown on the horizontal axis.

Fig. 4. Microbiota change during the fermentation of yamahai-moto.

Notes: Relative amounts of 16S rDNA contained in each yamahai-moto sample were calculated by qPCR (A). The results are shown as the relative abundances when the value of 1 day (early stage) was set to 1 (log relative abundance of 0). qPCR analysis was conducted in duplicate for each fermentation sample, and gave essentially the same results. Mean values are indicated in the figure. Taxonomic classification of pyrosequencing reads at the genus level are shown in (B), and Lactobacillus contents are indicated at the species level in (C). Genera or species are indicated in right insets. Genus classification using the RDP classifier was based on Table 2, and species classification was based on BLAST.

Change of total bacterial population during yamahai-moto fermentation

The qPCR analysis of 16S rDNA indicated a drastic increase in total bacteria 1–6 days from the starting time, corresponding to ~3 on a logarithmic scale (equivalent to a 1000-fold increase, Fig. 4(A)). This is consistent with the result of viable counts, which increased from approximately 10⁵–10⁶ to 10⁸ cfu/g during the period (Fig. 2(A)). The population then gradually decreased, but L. sakei remained dominant until 17 day (Fig. 4(C)). This result indicates that the vigorous growth of L. sakei during the first 6 days is critical in the fermentation of yamahai-moto.

Isolation of bacterial strains from yamahai-moto

We next analyzed the 16S rDNA sequences of bacterial strains isolated on solid MRS medium to identify their species (Table 3). Among the 139 strains, the early stage (0–5 days) isolates were predominantly Staphylococcus gallinarum (29 of 43 strains). Other strains were L. sakei (11 strains) and Leuconostoc citreum (3 strains). No Staphylococcus and Leuconostoc strains were isolated at the middle stage (6–11 days); strains grown from the middle and late stage (12–17 days) were solely identified as L. sakei (48 strains for each stage). These results supported those obtained in the pyrosequence survey, which confirmed the rigorous growth of L. sakei until 6 days. All of these members exhibited more than 99% identity with their respective type strains, L. sakei ATCC 15521 (GenBank accession number AM113784), L. citreum ATCC 49370 (AF111948), and S. gallinarum ATCC 35539 (D83366). The sequences of each isolate shared > 99% identity with the pyrosequence reads of the corresponding taxons. Note that we did not identify isolates of L. acidipiscis at any time points. We were unable to discover why this was, but suppose that it may have resulted from culture-dependent approach, which is greatly affected by the growth bias and the limitation of isolate numbers.

Table 3. Taxonomic classification of bacterial strains isolated from yamahai-moto, based on 16S rDNA.

Fermentation stage	Species identified (number of isolates)^a
Early stage (0–5 days after the start of fermentation)	Staphylococcus gallinarum (29), Leuconostoc citreum (3), Lactobacillus sakei (11)
Middle stage (6–11 days)	L. sakei (48)
Late stage (12–17 days)	L. sakei (48)

^a Species identified by BLAST.

Discussion

This study demonstrated the energetic growth of L. sakei in yamahai-moto fermentation starter. Yamahai-moto fermentation has been regarded as a well-regulated process: (i) the enzymatic actions of A. oryzae contained in kome-koji provide the necessary nutrients, linoleic acid and peptide-forming arginine, to stimulate the growth of L. sakei,³³⁾ (ii) nitrate-reducing bacteria form appropriate concentrations of nitrite to repress the growth of wild yeasts, film-forming yeasts, and other putrefactive bacteria in the starting materials.^{2, 34–36)} Along with these events, undesirable micro-organisms gradually decrease their population and Leuconostoc and L. sakei subsequently grow to provide an appropriate acidic condition that is suitable for alcohol fermentation by S. cerevisiae.³⁷⁾ Therefore, initial-stage microbial succession is very important in the yamahai-moto fermentation, as this significantly influences the stability and quality of yamahai-moto.

The microbiota revealed in this study, however, showed a quite different succession; L. acidipiscis was present in the early stage, in the absence of nitrate-reducing bacteria (Table 2 and Fig. 4). Actually, we detected little nitrite during yamahai-moto fermentation (less than 0.1 ppm), although considerable amounts have been reported in other studies (up to 27 ppm).^{34, 38)} This indicates that yamahai-moto fermentation has multiple microbiota types, which may be differently controlled with or without nitrite synthesis. It is known that failure in the control of initial microbiota results in a hasty abnormal fermentation, called haya-waki (putrefactive and incomplete maturation produced by wild yeasts)³⁹⁾; perhaps the presence of lactobacilli, possibly L. acidipiscis, in the early stage could prevent putrefaction when nitrites are absent. The 16S rDNA sequence of L. acidipiscis (type strain FS60-1, GenBank accession number AB023836) is largely different from those of the type strains L. sakei ATCC 15521 (AM113784) and L. plantarum ATCC 14917 (D79210) (88% identity with each), therefore it is improbable that the presence of L. acidipiscis in yamahai-moto has been missing throughout the past reports because of the similarity of the 16S ribotype.

Several recent reports demonstrate that the micro-organisms included in the moto fermentation are indigenously regulated, depending on the brewery’s process environment. Bokulich et al.¹²⁾ suggested, after using a marker gene quantitative PCR approach, that micro-organisms inhabiting the sake brewery’s room surfaces or cellar equipment could essentially influence the fermentation brew microflora. In another report, Masuda et al.¹¹⁾ characterized multiple isolates from a sake fermentation tank surface and moto using pulsed-field gel electrophoresis, and revealed that identical strains of L. mesenteroides were present over several years, and stably grew in that brewery’s moto. Considering these reports, L. acidipiscis and other early stage Lactobacillales detected might have resided in fermentation equipment or on the cellar surface we prepared kome-koji in this study. We isolated a different Leuconostoc species, L. citreum, in this study; this species is known sometimes to replace L. mesenteroides in the moto fermentation.⁴⁰⁾ It would be interesting to determine whether different Leuconostoc strains inhabit different breweries, thereby establishing the unique flavor and taste of each yamahai-moto.

We previously reported that L. sakei occupied the major part of microbiota in Japanese medieval sushi, kaburazushi, which is made by fermenting salted turnip and salted yellow tail in the presence of kome-koji.²²⁾ It is therefore conceivable that L. sakei could originate from kome-koji, which is the only shared ingredient used without undergoing any bactericidal process among yamahai-moto and kaburazushi. Lactic acid bacteria contained in kome-koji could be quite low in number as investigated by Ashizawa and Saito (at not more than 10² /g),⁴¹⁾ and major bacteria in kome-koji were Bacillus and Staphylococcus as reported by Bokulich et al.¹²⁾ Despite the low initial population, the process of moto manufacturing stimulates the vigorous growth of lactic acid bacteria. Future studies should address how the moto lactic acid bacteria are differently controlled to become dominant at the species level, as with the fermentation of other kimoto¹²⁾ and Korean traditional rice wine takju⁴²⁾ which contains L. plantarum as its major constituent.

Taken together, this study illustrates that the early yamahai-moto fermentation stage is divided into the nitrite and non-nitrite types. Although the results obtained in this study should carefully be interpreted because they are derived from a case study in one specific brewery’s process environment, the observed microbiota succession indicated that yamahai-moto is more diversely regulated than previously understood.

Author contributions

Conceived and designed the experiments: Takashi Koyanagi, Masashi Kiyohara, Hiroshi Matsui, Takane Katayama, and Hidehiko Kumagai. Performed the experiments: Takashi Koyanagi, Masashi Kiyohara, and Hiroshi Matsui. Analyzed the data: Takashi Koyanagi, Akira Nakagawa, Masashi Kiyohara, and Takane Katayama. Contributed reagents/materials/analysis tools: Atsushi Tsuji, Florin Barla, Harumi Take, Yoko Katsuyama, Koji Tokuda, Shizuo Nakamura, Hiromichi Minami, Toshiki Enomoto. Wrote the paper: Takashi Koyanagi and Takane Katayama.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was supported by the City Area Program (Central and Northern Ishikawa Area, 2009) of the Ministry of Education, Culture, Sports, Science, and Technology, Japan.

Acknowledgments

We thank Dr. Jiro Nakayama and Dr. Kenji Sonomoto for their helpful advice on pyrosequencing analysis.