Culture-dependent and -independent investigations of bacterial migration into doenjang from its components meju and solar salt

We determined bacterial migration into doenjang from its components, meju and solar salt using culture-based and 16S rRNA gene-based culture-independent techniques (pyrosequencing of total DNA). Pyrosequencing results suggested that the bacterial communities of meju, but not solar salt, significantly affected those of doenjang communities. Culture-based pyrosequencing analysis yielded similar results. These results indicate that most predominant bacterial species in doenjang migrated from meju, not solar salt. We therefore believe that the present study is one of the most comprehensive comparisons of bacterial communities of fermented soybeans using culture-dependent and -independent methods. Furthermore, pyrosequencing of the V3 and V4 regions of bacterial 16S rRNA did not distinguish among Bacillus amyloliquefaciens, B. siamensis, and B. velezensis as well as between Enterococcus faecium and E. hirae.


Introduction
Doenjang is a traditional Korean high-salt-fermented soybean paste used as a sauce, ingredient, or both for a variety of foods because of its nutritional and sensory properties [1]. The microbial communities of doenjang confer its quality and flavor [2]. Culture-dependent results suggested that most bacteria involved in doenjang fermentation are derived from raw materials such as its major component meju (fermented soybean block), which is prepared from soybeans by soaking, steaming, mashing, and molding, followed by fermentation for one or two months under natural environmental conditions [1]. Doenjang prepared from ripened meju contains approximately 18% high-salt brine [3,4]. The salt contents of meju and doenjang are significantly different (>10% w/v), and Jeong et al [5] suggested that difference of salts affects the compositions of microbial communities.
Recently, culture-independent methods such as pyrosequencing, which are increasingly applied to determine the composition of microbial communities in fermented soybean, have identified diverse microorganisms, including previously unidentified species that are not detected using culture-dependent methods [6][7][8][9][10]. However, the migration of the microbial community from meju to doenjang remains uncharacterized because it is difficult to obtain the same lots of meju because of the long fermentation process (>4 months) [5]. In our previous culture-dependent study to investigate bacterial migration from meju to doenjang, we identified Enterococcus and Tetragenococcus as the dominant genera in meju and doenjang from the Gyeonggi Province of Korea, respectively; and Bacillus is the dominant genus throughout the entire process [5]. These results indicate a shift in the salt concentration of meju (approximately 1.5%) to doenjang (approximately 12%) that is mediated by the dominant species. The predominance of salt-tolerant bacteria, including T. halophilus, suggests their presence in solar salt and their proliferation during the fermentation of doenjang. However, we are unaware of studies designed to identify the predominant species of doenjang derived from solar salt. Here we investigated bacterial migration using culture-independent and -dependent analyses. These results delineate the characteristics of bacterial migration into doenjang from its components and suggest the involvement of diverse bacterial species.

Meju, solar salt, and doenjang
Meju, solar salt, and doenjang samples were purchased from a traditional manufacturer in the Seosan area of Korea on different timetables [11]. Meju, fermented soybean for 2 months, and solar salt were used materials of purchased doenjang which was fermented for 3 months. Samples were ground, homogenized, or both, with an equal amount of sterilized water, then filtered through sterilized cheesecloth. The filtrates were analyzed for NaCl concentration, pH values, bacterial compositions, and for their culture-independent bacterial communities. NaCl concentration was measured using silver nitrate according to the Mohr method [12], and pH was measured using a pH meter. Microbial counts were determined by spreading appropriate dilutions of the filtrates on Plate Count Agar (PCA) (BD Difco, Detroit, MI, USA), and PCA containing 7% or 12% NaCl as a final concentration.

Pyrosequencing of culture-independent samples
To perform a culture-independent metagenomics analysis of the bacterial community, a DNeasy PowerSoil Kit (Qiagen, Hilden, Germany) was used to prepare DNA from filtrates of meju, solar salt, and doenjang. DNA quantity and quality were determined using PicoGreen and a Nanodrop. The V3 to V4 hypervariable regions of the bacterial 16S rRNA gene from genomic DNA were amplified using the primers as follows: forward, 5 0 -TCG TCG GCA GCG TCA GAT GTG TAT AAG AGA CAG CCT ACG GGN GGC WGC AG -3 0 and reverse, 5 0 -GTC TCG TGG GCT CGG AGA TGT GTA TAA GAG ACA GGA CTA CHV GGG TAT CTA ATC C -3 0 ) using a MyCycler Thermal Cycler (Bio-Rad, USA). PCR mixtures contained 30 ng of genomic DNA, 50 pM each primer, and Han-Taq polymerase (Genenmed, Korea). PCR was performed as follows: initial denaturation at 94˚C for 90 s, 30 cycles of denaturation at 94˚C for 45 s, annealing at 55˚C for 45 s, extension at 72˚C for 45 s, and final extension at 72˚C for 5 min.
Each primer was concatenated to an 8-base sample-specific barcode sequence and a common linker sequence (TC, forward primer and CA, reverse primer) at the 5 0 terminus [13]. Pyrosequencing was performed by Macrogen (Korea) using the MiSeq platform (Illumina, San Diego, USA) according to the manufacturer's instructions. To assess richness estimators of bacterial species, diversity indices, and rarefaction curves, we applied the pyrosequencing pipeline of the Ribosomal RNA Database Project [14]. Venn diagrams were generated using the packages gplots and limma in R v.4.0.0 (http://www.R-projet.org).

Identification of culture-dependent isolates using 16S rRNA gene sequence analysis
Culture media used to isolate bacteria were as follows: MRS agar (BD Difco), MRS agar supplemented with 7% or 10% NaCl, Nutrient Agar (BD Difco), and Nutrient Agar containing 7% or 12% NaCl. All media were incubated at 30˚C until distinguishable colonies appeared, and >20 types of colonies were collected from each plate according to differences in morphology, growth characteristics, and the number of colonies. The colonies were purified through successive transfers to plates containing the same type of agar medium used for isolation.
Genomic DNAs of isolates were extracted using a DNeasy Blood & Tissue Kit (Qiagen, Germany). Amplification of the 16S rRNA gene was performed using eubacterial universal primers 27F (5 0 -AGA GTT TGA TCC TGG CTC AG-3 0 ) and 1492R (5 0 -GGT TAC CTT GTT ACG ACT T-3 0 ) [15] with a T3000 Thermocycler (Biometra, Germany). The PCR mixture comprised template DNA (10 ng), 0.5 mM each primer, 1 U of Taq polymerase (ImClone, Daejeon, Korea), 10 mM dNTPs, and 2.5 mM MgCl 2 . Samples were heated for 5 min at 95˚C and then amplified as follows: 30 cycles for 1 min at 95˚C, 1 min at 58˚C, and 1 min at 72˚C. The PCR products were purified and sequenced using a custom service provided by Bionics Co. (Seoul, Korea). We used a web-hosted BLASTn algorithm to query the National Center for Biotechnology Information database for 16S rRNA gene sequences (http://blast.ncbi.nlm.nih. gov). The phylogenetic relationships of the isolates were inferred according to this analysis.

Physicochemical analysis and viable counts
Meju and doenjang samples had average NaCl concentrations of 2.1%, and 13.9%, respectively ( Table 1). The salt contents of meju and doenjang were similar to those acquired from other regions [5], but differed in pH values. The salt concentrations and pH values of solar salt preparations vary [16], and in the present study, the salt content was 69.2% (pH 4.87).
The average concentrations of the bacterial populations of meju and doenjang, determined using PCA, were 7.8 × 10 8 CFU/g and 6.7 × 10 8 CFU/g, respectively, and 1.1 × 10 3 CFU/g and 7.5 × 10 3 CFU/g using PCA supplemented with 12% NaCl ( Table 2). The NaCl concentration of doenjang was approximately 12%, thus these results suggest that the high NaCl concentration of doenjang inhibited bacterial growth during fermentation. In solar salt, there were few viable bacteria (averages of 4.3 × 10 1 CFU/g and 2.5×10 0 CFU/g on PCA and PCA with 12% NaCl, respectively). The cell counts were greatly influenced by the NaCl concentrations of the plating media.

Bacterial communities identified using culture-independent analysis
Analyses of meju, solar salt, and doenjang yielded 190,634; 202,956; and 123,972 sequences of sufficient quality, respectively. The sequence coverage of these samples ranged from 0.99 to 1.00, which provided sufficient statistical power to conduct analyses of bacterial communities. The numbers of operational taxonomic units and Chao1 (species richness), and Shannon indexes (species diversity) are shown in Table 3. The Chao1 and Shannon indexes ranged from 30-351 and 0.65-5.68, respectively, indicating that the species in solar salt were more diverse than those of doenjang and meju.
The species pools shared by the bacterial communities in meju, solar salt, and doenjang are depicted in Fig 2A. Three samples include B. velezensis and Erwinia aphidicola, which accounted for 89.59%, 0.05%, 64.51% of the species in meju, solar salt, and doenjang,

PLOS ONE
respectively. Except for two species, the others were not shared between solar salt and doenjang. These results indicate that solar salt had an insignificant effect on the composition of the bacterial communities in doenjang. Notably the seven species shared between meju and doenjang accounted for 98.88% and 87.62% of their populations, respectively. These results indicate that most bacterial communities of meju populated doenjang.

Bacterial communities identified using culture-dependent analysis
The most frequently identified bacterial species in solar salt were not consistent with those of meju and doenjang. In addition, viable bacteria from solar salt did not obtained on nutrient and MRS by culture-dependent analysis. Therefore, we conducted culture-dependent approaches, which identified 130 and 123 species in meju and doenjang samples, respectively (Table 4). Bacilli formed colonies on nutrient and MRS agar independent of NaCl concentration, and lactic acid bacteria (LAB) were isolated using MRS agar. The addition of NaCl to MRS decreased the numbers of LAB, E. faecium, and Leuconostoc mesenteroides, while halotolerant LAB and T. halophilius formed colonies on MRS containing 10% NaCl. Furthermore, eight species of Bacillus; the LABs E. faecium, L. mesenteroides, and T. halophilus; and one staphylococcal were identified. As the NaCl concentration increased, changed from E. faecium to T. halophilus, while the number of bacillus species did not change. Bacillus species were the most abundant in meju and doenjang, and the population of B. velezensis was the largest.
E. faecium, which was the predominant species in meju, was not isolated from doenjang and did not grow in medium containing 7% NaCl. These data agree with a previous report that the optimum concentration of NaCl for growth of Enterococcus is 6.5% [17]. Therefore, these results support the conclusion that the high salt concentration of doenjang inhibited the growth of E. faecium.
T. halophilus was isolated from doenjang and meju using media containing NaCl. Other investigators suggest that the halophile T. halophilus migrates into doenjang from solar salt during the brining step [9,10]. Conversely, our culture-independent method did not detect T. halophilus in solar salt (S1 Table), and T. halophilus was isolated from meju and doenjang using culture-dependent methods (Table 4). Therefore, we carefully suggest that T. halophilus migrated from meju to doenjang. S. saprophyticus was isolated here only from meju using media containing NaCl, although meju samples contain 2.1% NaCl. Our previous studies found that coagulase-negative staphylococci cultures containing S. saprophyticus are predominantly detected in meju and doenjang and contribute to its enhanced sensory properties [18][19][20]. However, only S. saprophyticus was detected in the present study.
Here we show that the species shared by the bacterial communities of meju, and doenjang were as follows: B. amyloliquefaciens, B. siamensis, B. subtilis, B. velezensis, and T. halophilus  (Fig 2B). These five shared species accounted for 80.00%, and 96.75% of those in meju and doenjang, respectively. However, previous our culture-dependent studies showed that shared species and ratio were diverse [5]. High shared rate of species in current study should be derived from salt-tolerance B. velezensis, which species were detected dominantly in meju, while salt-sensitive species such as Enterococcus faecium were detected dominantly from meju in previous study and thus those species might be not survived in high-salt condition, doenjang [5]. Consequently, culture-dependent results revealed that the bacterial communities of doenjang were derived from meju and that the high-salt concentration of doenjang accounted for the differences in their bacterial populations.

Comparison of bacterial communities identified using culture-dependent and culture-independent analyses
Culture-independent analysis showed greater diversity in bacterial communities compared with culture-dependent analysis (Fig 2C and 2D). Between two methods, although the number of shared bacterial species was low, the portions of shared species were high. For example, in meju, 50 and 8 species were detected using culture-independent and -dependent analyses, respectively (Fig 2C), and only one species (B. velezensis) was identified using both techniques, which accounted for 89.55% and 66.15% of the analyses, respectively. Eight species were uniquely detected using the culture-dependent technique, and four of eight species were B. amyloliquefaciens, B. siamensis, B. subtilis, and B. velezensis, which share >99% identities among their 16S rRNA gene sequences. Interestingly, the sequences of the V3 to V4 hypervariable regions of did not distinguish B. amyloliquefaciens, B. siamensis, and B. velezensis from each other, compared with those of the entire 16S rRNA gene (Fig 3). Similarly, E. faecium and E. hirae were not distinguished according to the sequences of their V3 to V4 hypervariable regions. We assumed that although B. velezensis was annotated according to the pyrosequencing data, the population may change from B. velezensis to B. amyloliquefaciens or B. siamensis as well as E. hirae, which annotated according to the pyrosequencing analysis, may be replaced by E. faecium.
Here we identified 26 and 9 species in doenjang using culture-independent and culturedependent analyses, respectively ( Fig 2D); and B. velezensis and T. halophilus were identified by each. These species accounted for 62.66% and 85.27% of those identified by culture-independent and -dependent analysis, respectively. Eight bacilli were detected using the culturedependent assay, while only one was detected using the culture-independent assay. These findings indicate that culture-dependent analysis was more precise and that the findings of culture-independent analysis and pyrosequencing included those of culture-dependent analysis.
The present study reveals the advantages and limitations of the two types of analyses. The former serves as a powerful tool to evaluate microbial diversity. Although the latter method identifies viable cells, which are used to screen for starter candidates, it does not sufficiently assess bacterial diversity. Therefore, the complementary results using the two methods here may likely be explained by bacterial migration from meju into doenjang. Our present results show that most predominant bacterial species of meju may migrate into doenjang during fermentation.
In the present study, the culture-independent method most frequently identified L. lactis in meju and doenjang. However, L. lactis was not isolated using culture-dependent methods. Leuconostoc is the dominant genus of kimchi, the traditional Korean fermented kimchi cabbage, which is used as a commercial starter for kimchi fermentation [33]. However, this genus is rarely detected as a minor group in high-salt fermented foods such as doenjang and jeotgal using culture-dependent approaches. Culture-independent analyses such as PCR-denaturing gradient gel electrophoresis found that the LAB population, including Leuconostoc, is dominant in meju and doenjang [7,34,35]. However, we show here that most LAB were not sufficiently salt-tolerant to proliferate in doenjang. Furthermore, we did not detect E. faecium and L. mesenteroides using media containing NaCl (Table 4), consistent with the results of a previous study [5]. Therefore, culture and pyrosequencing results indicate that LAB may be present in meju.
We believe that the present study is the most comprehensive comparison of culture-independent and -dependent methods used to track bacterial migration from meju into doenjang. Although the bacterial populations identified using culture-dependent analysis were less diverse than with the culture-independent analysis, the latter identified the predominant bacterial species. This suggests that novel culturing techniques are required to identify process-relevant bacteria and to develop starter candidates. Although the use of pyrosequencing wildly outpaces culture-dependent analysis, more accurate species identification must be achieved. Moreover, the complementary nature of two methods should increase our understanding of bacterial migration and identify novel bacteria that are present in high-salt fermented foods.
Supporting information S1