Doenjang is a Korean traditional soybean paste popularly consumed as a condiment for vegetables, fish, and meats or used as a seasoning ingredient in authentic Korean cuisine. The paste has received considerable attention because of numerous reported beneficial human health effects, including antioxidant, fibrinolytic, antimutagenic, and anticancer properties (Kim, 2004; Yun, 2005; Jung et al., 2006; Park et al., 2008; Namgung et al., 2009; Kwon et al., 2010; Tamang et al., 2016a).

Culture-based approaches have been widely applied to bacterial community analysis of doenjang (Yoo et al., 1999; Jeong et al., 2014), but they have produced limited information because culturing is time-consuming and laborious, and because doenjang contains unculturable microbes. Recently, culture-independent methods, such as denaturing gradient gel electrophoresis (DGGE) and pyrosequencing, have been widely used to investigate bacterial communities in doenjang (Cho and Seo, 2007; Kim et al., 2009; Nam et al., 2012). However, previous studies using culture-independent methods have limited their analyses to snapshots of bacterial communities by focusing on short-time frames within the doenjang fermentation process. To the best of our knowledge, thus far, no study has been conducted to investigate microbial community fluctuation over the full doenjang fermentation period. In Korea, traditional doenjang is typically made by further fermentation of the solid parts from a fermented mixture of doenjang-meju (fermented soybean bricks) and brine. The additional fermenting procedure also suggests that the microbial community and indigenous enzymes in doenjang-meju are likely important in determining the microbial community and metabolite change during doenjang fermentation. However, no research exists on how doenjang microbial communities alter when doenjang-meju with known microbial community composition is used.

Traditional doenjang is produced by spontaneous fermentation without the use of starter cultures, leading to the growth of diverse microorganisms. In turn, quality variation of doenjang products tends to result, as well as the occasional production of undesirable metabolites, such as biogenic amines (BAs) or toxins (Cho and Seo, 2007; Shukla et al., 2010; Park et al., 2014). Most previous studies have focused on the analysis of either microbial communities or metabolites in doenjang (Cho and Seo, 2007; Kim et al., 2009; Rhyu and Kim, 2011; Nam et al., 2012), which makes it difficult to investigate microbial functional properties during doenjang fermentation. Instead, examining microbial successions and metabolite changes simultaneously is crucial for a better understanding of microbial community function in doenjang. However, such studies have not yet been performed.

Pyrosequencing based on 16S rRNA gene sequences has been broadly applied to analyze microbial communities in fermented foods, because it yields more detailed data compared with conventional microbiological methods, such as DGGE and culture-based approaches (Nam et al., 2012; Park et al., 2012). Additionally, proton nuclear magnetic resonance (¹H NMR) spectroscopy is one of the easiest, yet most comprehensive, and powerful tools to analyze diverse metabolites simultaneously in fermented foods (Jeong et al., 2013; Jung et al., 2013; Lee et al., 2015). In this study, we used pyrosequencing and ¹H NMR techniques to investigate microbial succession and metabolite changes, respectively, across the full length of doenjang fermentation. The resultant data will increase our knowledge regarding the functional properties of major microbial communities involved in doenjang fermentation.

Traditional doenjang is produced by a long-term fermentation of solid parts obtained from a doenjang-meju and brine mixture, which is prepared by soaking the fermented doenjang-meju bricks in an 18–20% (w/v) solar salt solution for 40–60 days without the use of starter cultures. Because both bacteria and fungi are involved in doenjang-meju fermentation (Kim et al., 2010b; Lee J.H. et al., 2010; Jung et al., 2014) and have been detected in doenjang samples (Kim et al., 2009; Shukla et al., 2014), researchers have suggested that the two microbial groups may also play important roles in doenjang fermentation. We noted the presence of both bacteria and fungi in our doenjang samples (Figure 1). However, viable bacterial cell counts were approximately three orders of magnitude higher than fungal counts during fermentation. In addition, doenjang fermentation mostly occurs under anaerobic and high moisture conditions. Thus, fungi that prefer dry and aerobic conditions may exhibit lower metabolic activity compared with bacteria during this process. We therefore inferred that the observed fungal growth derived mostly from spores without metabolic activity that had originated in the doenjang-meju. This inference was supported by the lack of hyphae or mycelia in the doenjang samples (data not shown). These results suggest that bacteria are probably more responsible for doenjang fermentation than fungi. Hence, we chose to focus on bacteria in the microbial community analysis.

Although 16S rRNA gene sequencing is the most powerful tool in bacterial taxonomy, it is limitedly used at the species level classification due to the low phylogenetic resolution and poor discriminatory power of 16S rRNA gene sequence at the species level (Fox et al., 1992; Janda and Abbott, 2007). It has been generally accepted that it would be impossible to assign 16S rRNA gene sequences to the level of species because the taxonomies end at the level of genus and DNA relatedness studies are necessary to provide absolute resolution to these taxonomic problems (Schloss and Westcott, 2011). Therefore, in this study we taxonomically classified the 16S rRNA gene sequencing reads at the phylum and genus levels. The genera observed on day 0 of fermentation [Bacillus, Enterococcus, Lactobacillus, Clostridium, Staphylococcus, Corynebacterium, and Oceanobacillus (Figure 3B)] accorded well with genera detected in previous snapshot analyses of doenjang fermentation (Cho and Seo, 2007; Kim et al., 2009; Nam et al., 2012; Jeong et al., 2014). With the exception of Oceanobacillus, all genera had also been identified from doenjang-meju (Jung et al., 2014), despite the preparation step of soaking the bricks in solar salt solution for 42 days. These results suggest that doenjang-meju is a major source of bacteria in doenjang fermentation. In contrast, Oceanobacillus and Tetragenococcus (a dominant group at 108 days of fermentation) likely derived from solar salts because they were not found in doenjang-meju (Jung et al., 2014). Moreover, the two genera are halotolerant or halophilic groups that have been frequently isolated from high-saline environments, including solar salterns (Baati et al., 2010; Lee S.Y. et al., 2010). Thus, solar salts appear to be another important bacterial source in doenjang fermentation.

Bacillus was one of the predominant populations in doenjang during the entire fermentation period (Figure 3B). Due to the well-documented predominance of Bacillus in both culture-dependent and culture-independent studies on doenjang, researchers have generally assumed that this group is primarily responsible for the fermentation process (Yoo et al., 1999; Kim et al., 2010a; Jeon et al., 2016; Tamang et al., 2016b). However, members of Bacillus grow aerobically, and therefore, we consider it unlikely that they can be principal actors in doenjang fermentation, which occurs almost entirely under anaerobic conditions. Furthermore, most Bacillus members do not grow well under conditions more saline than 15% NaCl, although they have been detected from doenjang samples with ∼18% salt concentrations (Jung et al., 2014, 2016). Our data also showed that Bacillus abundance was relatively constant throughout fermentation and was unassociated with fluctuations in metabolite content (Figure 4). Therefore, we infer that Bacillus in doenjang are probably metabolically inactive or exist as spores. Although their relative abundance is high, we infer that this genus is not primarily responsible for doenjang fermentation.

Enterococcus are facultative anaerobic, Gram-positive, cocci-shaped lactic acid bacteria that occurred dominantly alongside Bacillus during the early fermentation period (Figure 3B). This finding corresponds to previous studies that have also reported Enterococcus as a dominant population in doenjang (Cho and Seo, 2007; Kim et al., 2009; Nam et al., 2012). Some members of Enterococcus are pathogenic and have caused infections in humans (Fisher and Phillips, 2009), but other species such as E. faecalis or E. faecium have been frequently isolated from fermented foods and even used as probiotics (M’hir et al., 2012; Todorov and Holzapfel, 2015). Therefore, the dominance of Enterococcus was expected and likely consisted of non-pathogenic species. The high counts of Enterococcus during the early fermentation period was associated with the increase of glucose, fructose, galactose, and glycerol (Figures 3B and 4A), which were probably generated through the hydrolysis of polysaccharides, flavonoid glycosides, and lipids in doenjang-meju. However, most Enterococcus members do not show glyosidic and lipase activities under high saline conditions such as doenjang (Jeong et al., 2014), suggesting that their metabolism was too low during fermentation to hydrolyze the organic carbon compounds. Instead, we speculated that endogenous enzymes derived from doenjang-meju might be responsible for the observed increases in glucose, fructose, galactose, and glycerol. Moreover, despite the dominance of Enterococcus, we only observed slight increases in lactate and acetate during early fermentation, a finding that supports the hypothesis of weak metabolic activity in these lactic acid bacteria (Figures 3B and 4B). However, additional studies will be necessary to investigate how organic carbon compounds increase in doenjang during the early fermentation period.

Bacterial community analysis showed that Staphylococcus increased to approximately 7.6% of the total bacterial abundance at 28 days of fermentation (Figure 3B). Similar to Enterococcus, some members of this genus are pathogenic, but other Staphylococcus species are considered starters for fermentation and accordingly, have been identified from several fermented foods, including doenjang, sausage, and fish sauce (Fontana et al., 2005; Yongsawatdigul et al., 2007; Kim et al., 2009; Guan et al., 2011; Jung et al., 2013; Lee et al., 2015). Therefore, the Staphylococcus species observed in our doenjang samples probably do not have pathogenic properties, but additional research is required to understand their exact role during fermentation, as the function of Staphylococcus was unclear in this study.

Bacterial community analysis demonstrated that Tetragenococcus, a genus of halophilic, Gram-positive lactic acid bacteria, was a dominant group in 108-day samples (Figure 3B). The dominance of Tetragenococcus corresponded well to the rapid increases in lactate and acetate concentrations, as well as the decreases in glucose and fructose, occurring around the same time (Figures 3B and 4A,B). Additionally, the drop in pH at 108 days was in line with heightened lactate and acetate production (Figure 1). Together, these results suggest that Tetragenococcus may be primarily responsible for the production of organic acids during doenjang fermentation. BAs, including putrescine, cadaverine, spermidine, histamine, and tyramine, are low-molecular-weight nitrogenous organic compounds produced via microbial decarboxylation of amino acids and nitrogen compounds during food fermentation (Shalaby, 1996; Park et al., 2000; Moon et al., 2010; Costantini et al., 2013; Jung et al., 2015). These compounds are known to be produced in doenjang (Shukla et al., 2010; Kim and Ji, 2015), but the agents of their production are unknown. The increase of putrescine and tyramine, respectively, generated through ornithine and tyrosine decarboxylation, was also associated with Tetragenococcus dominance (Figures 3B and 4C). These results suggest that Tetragenococcus may be a key agent in biogenic-amine production during doenjang fermentation. Previous studies have also shown that members of Tetragenococcus play important roles in the fermentation of salted goods, including fermented seafood; thus, this genus may be a good bacterial starter for flavor enhancement of fermented foods (Chen et al., 2006; Udomsil et al., 2011; Kim and Park, 2014; Jung et al., 2015, 2016). However, some Tetragenococcus species appear to produce BAs primarily via plasmid-encoded decarboxylation genes (Satomi et al., 2008, 2011, 2014), which increases the difficulty of using Tetragenococcus as starters for salted food fermentation. To address this problem, strains without the ability to produce these compounds have been applied to control biogenic-amine generation during fermentation (Udomsil et al., 2011; Kuda et al., 2012).

Interestingly, Lactobacillus, a group of Gram-positive, facultative anaerobic or microaerophilic, rod-shaped lactic acid bacteria, was identified as a dominant group in 179-day samples (Figure 3B). The non-protein amino acid GABA is a major inhibitory neurotransmitter and is produced through the irreversible α-decarboxylation of L-glutamate by glutamate decarboxylase. Lactobacillus species have been reported as major GABA-producing bacteria during fermentation (Makarova et al., 2006; Cho et al., 2007; Park and Oh, 2007; Siragusa et al., 2007; Komatsuzaki et al., 2008; Jeong et al., 2013). Our metabolite analysis supported these previous findings; galactose decreases and GABA increases were associated with the increase of Lactobacillus (Figures 4A,D), implying that members of Lactobacillus are responsible for GABA production during fermentation. Additionally, the ability of Lactobacillus to metabolize galactose is well-established and members of Lactobacillus have been detected in doenjang (Cho and Seo, 2007; Nam et al., 2012), indicating that lactate production from galactose and GABA synthesis by Lactobacillus are important processes during doenjang fermentation although no report showing their growth in 18% NaCl conditions exists.

To the best of our knowledge, this was the first study to investigate fluctuations in microbial communities and metabolite production simultaneously throughout the entire doenjang fermentation period. Ours was also the first to use doenjang-meju of known bacterial community composition in a study of doenjang fermentation. Here, we suggested that both doenjang-meju and solar salts are important bacterial sources in doenjang fermentation. Furthermore, we proposed that despite their overall abundance, Bacillus may be not as central to doenjang fermentation as previously assumed. Additionally, we showed that solar-salt-derived Tetragenococcus appears to be a primary producer of organic acids and BAs during doenjang fermentation, suggesting that Tetragenococcus strains without this ability are usable as starters, in order to reduce biogenic-amine concentrations. Finally, our results suggested that Lactobacillus is probably one of the major GABA producers during doenjang fermentation. In conclusion, this study contributed to an improved understanding of the biochemical processes underlying doenjang fermentation through exploring the functional properties of major doenjang microbial communities. The data generated should pave the way for additional research employing “omics” technologies, including metagenomics, metatranscriptomics, and metabolomics, which are certain to provide further insights into the production of safe and high-quality doenjang.

CJ conceived the ideas and supervised the work WJ and JJ developed the concepts and performed the experiments WJ and HL analyzed the data and CJ and WJ wrote the manuscript. The manuscript has been reviewed and edited by all authors.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.