Comparison of microbial diversity and metabolites on household and commercial doenjang

Highlights • Tetragenococcus genus among bacteria was identified as predominantly only in CDJ.• Zygosaccharomyces genus among yeast was identified as main in HDJ and CDJ.• Glutamic acid, associated with a unique umami taste, was predominated in HDJ and CDJ.• The 3-methyl butanal and benzeneacetaldehyde among VFCs were high in of HDJ and CDJ.• The isoflavone-aglycones, TFs, and enzymatic inhibitory activity were higher in HDJ.


Introduction
In Korea, doenjang, one of the most representative soybeanfermented foods in Asia, is generally divided into two types: homemade and commercially manufactured.(Jeong et al., 2017;Lee et al., 2017).Due to changes in socioeconomic position, family structure, and the proportion of working women, the industrial production of doenjang has dramatically increased over the past three decades (Jeong et al., 2017).The commercial doenjang (CDJ) processing time is much shorter (1 to 3 months) than the traditional household doenjang (HDJ) processing time (more than 6 months).CDJ is made by combining soybean koji (commercial meju), which is a culture of Aspergillus oryzae and Bacillus velezensis on soybean (Han et al., 2023;Zhang et al., 2022), with wheat koji that has been infected with A. oryzae and Aspergillus sojae (Li et al., 2023) prior to a brief fermentation time (3 to 7 days).Following the addition of soybean koji, wheat koji, soybean, and salted brine (10 to 14 %), the CDJ is fermented for 30 days without aging (Jeong et al., 2017).In contrast, HDJ is prepared from 100 % soybeans that are allowed to undergo spontaneous fermentation to produce traditional meju.In the winter, the meju is fermented outside, and in the spring, it is dried in the sun.At this time, various molds, bacteria, and yeast grow on meju.The meju undergoes a brining process in which salt water (10 to 14 %) is added, followed by fermentation for 60 days.After that, this fermentation product was divided into solid and liquid parts, and generally, the solid parts were again fermented (aging for about a year) to produce the HDJ (Jeong et al., 2017).Therefore, compared to CDJ, it takes a lot of time and is difficult to industrialize due to uneven quality.
Soybeans, the primary ingredients in doenjang, are rich in physiologically active biomolecules, including isoflavones, polyphenols, flavonoids, soy oligosaccharides, saponins, and lecithin that may be fortified by the microbes used during fermentation (Chen, Ni, Zhang, Liang, & Fang, 2023;Lee, Seo, & Lee, 2023).During fermentation, microbes hydrolyzed the main components, including proteins, lipids, carbohydrates, and flavonoid glycosides.At that time, several metabolites were created, including amino acids, organic acids, active metabolites, and aglycones, these metabolites increase the nutritional value of donejang, and decomposed soy proteins gave a distinctive flavor to doenjang (Zhang, Zhang, Guo, Ye, Zhao, & Huang, 2023) Previous studies, it is reported that the Aspergillus population was linked to sugar metabolism, Bacillus spp. was associated with fatty acids metabolism, and Tetragenococcus and Zygosaccharomyces were linked to amino acids metabolism (Lee et al., 2017).Therefore, the microbial of the doenjang plays a significant role in determining its chemical composition and quality.
Various microorganisms are not involved fermentation for CDJ.This is because the culture supported using a "starter" to speed up the fermentation process in a controlled environment (Lee, Yun, Lee, & Hong, 2022).Chun et al. (2020) reported that Bacillus and Aspergillus were the major strains involved in fermenting the traditional Korean doenjang.In addition to these strains, culture-independent methods showed that Mucor, Leuconostoc, Staphylococcus, Penicillium, Cladosporium, and Enterococcus play significant roles in the fermentation of meju (Ryu et al., 2021).Studies related to Korea's traditional doenjang and commercial doenjang include the correlation between microorganisms and metabolites in traditional soybean paste (Han, Chun, Kim, & Jeon, 2021) and a comparative evaluation of microorganisms and metabolites in soybean paste manufactured by two industrial manufacturers (Lee et al., 2017).It has not yet been adequately reported that comprehensively compared and analyzed the metabolites and microbial diversity of HDJ and CDJ.
Therefore, we collected four types of traditional HDJ from different cities and four types of CDJ Korea's representative industrial doenjang brand and analyzed the ability to inhibit α-glucosidase and pancreatic lipase, as well as their profiles of free amino acids (FAAs), isoflavones, and volatile flavor compounds (VFCs), for the industrialization and development of doenjang.In addition, by analyzing the bacterial and yeast diversity community that affects the quality of doenjang, we aim to provide basic information on microorganisms that can affect the quality of doenjang and apply this to the doenjang industry to contribute to the globalization of Korean doenjang.

Doenjang samples
In this study, eight samples of doenjang were used.Four types of HDJ (1HDJ, 2HDJ, 3HDJ, and 4HDJ) were collected from households in four cities in South Korea.The preparation process of HDJ and CDJ were detailed in Fig. S1.These four HDJ were commonly made using soybean, meju (fermented soybean block), solar salt, and water and fermented and aged for over a year.Four types of CDJ (1CDJ, 2CDJ, 3CDJ, and 4CDJ) were purchased from different manufacturers in South Korea.Ingredients of 1CDJ are composed of soybean, wheat flour, salt, meju, alcohol, roasting soybean, and fermentation microorganisms.Ingredients of 2CDJ are composed of soybean, wheat flour, salt, wheat rice, meju, L-monosodium glutamate, mustard powder, and fermentation microorganisms.Ingredients of 3CDJ are composed of soybean, wheat flour, salt, wheat rice, alcohol, seasoned yeast powder, mustard powder, and fermentation microorganisms.Ingredients of 4CDJ are composed of soybean, wheat flour, salt, wheat rice, alcohol, soybean flour, soybean fermentation concentrate, L-monosodium glutamate, and fermentation microorganisms.

Measurement of pH, acidity and salinity
pH, acidity and salinity in the doenjang samples were analyzed using the method by Lee et al. (2023).Each of the HDJ and CDJ samples (10 g) was shaken for 30 min with 90 mL of distilled water.The supernatant was then obtained by centrifugation (13,000 rpm, 3 min), and its pH, and acidity were measured.The pH was measured with a pH meter and acidity (lactic acid, %) was determined by titrating the supernatant with 0.1 NaOH until pH was 8.2 ± 0.1.This acidity was calculated using follow Eq.(1): Volume of titrate × N of titrate × Equivalent weight for lactic acid Weight of sample × 1000 (1) Equivalent weight for lactic acid: CH 3 -CHOH-COOH, M.W. = 90.For salinity measurements, each sample of HDJ and CDJ (10 g) was mixed with 40 mL of distilled water, and the supernatant was obtained after shaking and centrifugation, as stated above; its salinity (%) was calculated using a salt meter (PAL-03S, ATAGO, Japan).

Bacterial and yeast community analyses using unculture methods
Each 10 g of doenjang samples was mixed with 100 mL of phosphate buffered saline (pH 7.0) and shaken for 1 h.After filtering the samples through sterile gauze, the liquid supernatant was centrifuged at 13,000 rpm for 5 min at 4 • C. The G-spin genomic DNA purification kit was then used to isolate total DNA from pellet.Throughout the process, these kits were meticulously maintained following the manufacturer's recommendations.After that, the extracted DNA was used as a template for polymerase chain reaction (PCR) to amplify the 16S rRNA and 26S rRNA genes.The bacterial-specific PCR primer 5′-CGG AGA GTT TGA TCC H. Yul Lee et al.TGG-3′(1BF/18 mer, forward primer) and 5′-TAC GGC TAC CTT GTT ACG AC-3′ (1BR/20 mer, reverse primer) were used to amplify (40 cycles) 16S rRNA gene fragments (Cho et al., 2009a) and the yeast-specific PCR primers 5′-ACC CGC TGA AYT TAA GCA TAT-3′(3YF/21 mer, forward primer) and 5′-CTC CTT GGT CGT GTT TCA AGA CGG-3′(3YR/25 mer, reverse primer) were used to amplify (30 cycles) 26S rRNA gene fragments (Cho et al., 2009b).The 16S rRNA and 26S rRNA amplification were performed following methods described by Cho et al. (2009a) and Cho et al. (2009b), respectively.The PCR condition was followed by: initial denaturation at 95 ℃ for 5 min, 16S rRNA gene was followed by 40 cycles of denaturation at 94 ℃ for 30 sec, annealing at 49 ℃ for 30 sec, extension at 72 ℃ for 90 sec, and 26S rRNA gene was followed by 30 cycles of denaturation at 94 ℃ for 30 sec, annealing at 48 ℃ for 30 sec, extension at 72 ℃ for 60 sec after final extension at 95 ℃ for 5 min, the reaction was terminated by lowering the temperature to 4 ℃.The anticipated 16S rRNA and 26S rRNA PCR products of approximately 1.5 kb and 0.6 kb, respectively, were confirmed by electrophoresis on a 1 % agarose gel.After amplification, purification using MEGA quick-spin Total Fragment DNA Purification Kit (iNtRON Biotechnology Inc.) was performed.The purified genes were cloned in the pGEM-T easy vector (Promega, Madison, WI, USA) following the manufacturer's instructions and then transformed into E. coli DH5α competent cells (HIT competent cells T-DH5α, Real Biotech Corporation, Korea).Recombinant clones were picked randomly, and recombinant plasmids were extracted using the Plasmid DNA Purification Kit (iNtRON Biotechnology Inc.,).Nucleotide sequencing was analyzed by Genotech Corporation.The assembly of the nucleotide sequences was performed with the DNAMAN analysis system (Lynnon Biosoft, Quebec, Canada).All of the reference sequences were obtained from the National Center for Biotechnology Information (NCBI).The 16S and 26S rRNA sequence identity searches, sequences alignment, and the phylogenetic analysis were performed following the Cho et al. (2009a) method.

Measurement of FAA contents
The FAAs present in the doenjang samples were subjected to analysis utilizing the methodology delineated by Hwang et al. (2021).5 mL of distilled water was used to homogenize and hydrolyze 1 g of the sample at 60 • C for 1 h using a heating block (HB-48P, Daihan Scientific, Korea). 1 mL of 10 % 5-sulfosalicylic acid dihydrate was added after the mixture had been vortexed, and kept at 4 • C for 2 h.The liquid was then centrifuged for 3 min at 13,000 rpm after being filtered with a syringe filter.Using a rotary vacuum condenser, the filtrate was concentrated under reduced pressure at 50 • C, and then it was dissolved in 2 mL of lithium buffer (pH 2.2).The mixture was then quantitatively analyzed by an auto amino acid analyzer and filtered using a 0.45 μm membrane filter (Whatman, Maidstone, UK).

Analysis of VFCs
The confirmation of the VFCs in the doenjang samples was carried out by utilizing a modified version of the method previously reported by Cho et al. (2022).Using high vacuum sublimation, VFCs from doenjang samples were extracted using solid-phase microextraction (SPME) fibers (DVB/CAR/PDMS, 50/30 µM, Supelco-57329-U).A 20 mL headspace vial was filled precisely with 5 mL of the doenjang sample supernatant, sealed with an aluminium cap, and then adsorbed onto a 100 m polydimethylsiloxane (PDMS) fiber for 8 min at 100 rpm at 100 • C.Then, using an HP-5MS (30 m × 0.25 mm, 0.25 m film thickness) column, a GC-MS was used to examine the VFCs.The GC oven was heated from 40 • C to 180 • C at a rate of 5 • C/min, maintained at 180 • C for 3 min, and then the temperature was increased to 280 • C. Sample injection was done in splitless mode (10:1).The flow rate of the carrier gas, helium, was maintained at 1 mL/min.The quadruple and injector were both set to 110 • C temperatures.Electron ionization at 69.9 eV produced mass spectra with a mass scanning range of 50 to 550.To figure out what each VFC was, the mass spectrum of each peak was recorded from the chromatogram and compared to the GC-MS NITS library.

Preparation of extracts for TP, TF, and isoflavone contents and enzyme activities
The extracts were produced using a modified version of the methods proposed by Hwang et al. (2021).The 20 mL of 50 % ethanol (EtOH) was used to extract the powdered sample (1.0 g) over the course of 12 h at room temperature in a shaking incubator.Before HPLC analysis, the crude supernatant was centrifuged at 3,000g for 15 min.It was then filtered using a syringe filter (0.45 μm, Whatman, Maidstone, UK).These extracts were used for the analysis of TP, TF, and isoflavone contents.The extracts were lyophilized to prepare a powder after being concentrated in a rotary evaporator (N-1300, SHANGHAI EYELA CO., LTD, China) at 60 • C. To assess the enzyme inhibition activities, the powders were diluted to concentrations of 0.25, 0.5, and 1 mg/mL with 50 % EtOH.

Measurement of isoflavone contents
Isoflavone analysis was conducted using HPLC, following the method described by Hwang et al. (2021).The mobile phase involved using water (solution A) and acetonitrile (solution B) with 0.2 % glacial acetic acid.With solvent A, the analytical parameters were set at 100 %/0 min, 90 %/15 min, 80 %/25 min, 75 %/30 min, and 65 %/45 min.The solvent flow rate was then adjusted to 1 mL/min at 30 • C, and 20 μL of each sample was then injected.The absorbance at 254 nm was measured using a diode array detector (Agilent 1200 series, Agilent Co.).We established calibration curves by integrating the peak areas of isoflavones in the HPLC chromatogram at 254 nm and drawing a linear curve according to the concentration.The isoflavone stock solution was diluted with DMSO to achieve a concentration of 1,000 μg/mL, and calibration curves were sequentially prepared at eight different concentrations, including 0.5, 1, 5, 10, 20, 40, 80, and 100 μg/mL.In this process, we confirmed a high linearity with correlation coefficients (r 2 ) > 0.997 for each curve.

Measurement of TP and TF contents
The extracts were used to estimate TP and TF contents.The TP contents were calculated using the Folin-Denis method as outlined by Lee et al. (2018a), and the absorbance of the solution was measured at 750 nm.The TP contents were quantified utilizing the Eq. ( 2) procured from the gallic acid standard curve.
The method outlined by Lee et al. (2021) was used to determine the TF contents.The absorbance value was measured at 420 nm, and the TF contents was determined using the Eq. ( 3) derived from the rutin standard curve.TF = (Sample absorbance + 0.0013) ÷ 16.16 (3)

Evaluation of α-glucosidase and pancreatic lipase inhibition activities
The inhibitory activity of the extracts against α-glucosidase and pancreatic lipase enzymes was confirmed by slightly modifying the method previously reported by Hwang et al. (2021).The test tube was filled with 30 μL of each extract, 70 μL of either 1 U/mL α-glucosidase or of pancreatic lipase, and 50 μL of 200 mM sodium phosphate buffer (pH 6.8).Then, the mixture was pre-reacted at 37 • C for 10 min, and added to 100 μL of 10 mM p-NPG or p-NPB, dissolved in 200 mM sodium phosphate buffer (pH 6.8), and allowed to undergo reaction was reacted at 37 • C for 10 min.Finally, following the addition of 750 mL of 100 mM Na 2 CO 3 to halt the reaction, absorbance at 420 nm was measured.The following formula was used to compute the inhibitory activity of α-glucosidase and pancreatic lipase as a percentage (%) by the following Eq.( 4):

Statistical analysis
All data, except those related to bacterial and yeast communities, were expressed as a mean ± standard deviation.An analysis of variance (ANOVA) procedure followed by Duncan's multiple range test was conducted using the Statistical analysis system (SAS, version 9.4, SAS Institute, Cary NC, USA) software.The Duncan's multiple range test was used to confirm the significant variances among all treatments with a significance level of p < 0.05.The correlation between microorganisms and the concentrations of each major metabolites was evaluated by Pearson's correlation test and visualized with Graphad Prism 8.0.Meanwhile, when generating a heatmap of the correlation between microbial diversities and FAA or VFCs, the data were prepared by selecting mainly detected or specific substances in all doenjang samples.

Comparison of pH, acidity, and salinity in the HDJ and CDJ samples
As shown in supplementary Table S1 and Fig. 1, the pH values of HDJ and CDJ were in a weakly acidic range, varying from pH 5.14 to 5.94.Comparatively, the pH value of HDJ was higher than average values of doenjang samples.The acidity value was not significantly correlated with pH because, the pH value was lowest in 4CDJ (pH 5.14 and 2.87 %), while it but the acidity was highest in 4HDJ (pH 5.94 and 3.03 %), which had the highest pH.In terms of salinity, the value of HDJ was within approximately 10 %, at 8.20 to 10.20 %, and CDJ was 12.80 to 14.60 %, 2 to 4 % higher than HDJ.
HDJ samples exhibited a higher pH (5.94 to 5.31) than CDJ samples (5.64 to 5.14), consistent with previous results (Chun et al., 2020;Zhang et al., 2022).Han et al. (2021) and Ryu et al. (2021) reported that the pH value of doenjang increased during fermentation, while Tian et al. (2022) reported that it decreased.According to previous results, it is difficult to conclude that the fermentation period and pH of doenjang have a positive correlation.Because each report was different on pH changes during fermentation in previous research results.The difference in pH between HDJ and CDJ may be due to koji, which is the ingredient of CDJ, unlike HDJ.According to a study by Hong and Kim. (2020), it is reported to be the pH of rice koji and wheat koji was relatively low than soybean koji.Wheat koji is commonly used as a material for CDJ.On the other hand, HDJ was not used, and therefore, we think that CDJ has a relatively low pH due to this difference.Based on previous studies, the salinity concentration of doenjang is reported more than 11 % (11-13.61% (traditional) and 11.27-11.60% (commercial) (Ryu et al., 2021;Jang, Yu, & Kim, 2021).It is generally known that traditional doenjang has a higher salt content than commercial doenjang, but the results of this study showed the opposite trend.Unlike previous results, the HDJ salinity of this study was relatively low (Fig. 1).We believe that this is a characteristic of the HDJ production method used in the experiment.It is thought that the HDJ probably had a relatively large amount of salt eluted into the liquid (soy sauce) during the divide process of the doenjang.

Comparison of microbial diversities in the HDJ and CDJ samples 3.2.1. Bacterial diversity
The unculturable bacterial diversity assessd by 16S rRNA gene sequencing during the metagenomics analysis of the DNA of the HDJ and CDJ samples, is shown in Fig. 2A, and 2B, and supplementary Fig. S2 and  S3.The genera of the bacterial populations in HDJ and CDJ are depicted in Fig. 2A.There were 2, 2, 7, and 6 bacterial genera in 1HDJ, 2HDJ, 3HDJ, and 4HDJ, respectively.The predominant genus in 1HDJ and 2HDJ was Bacillus (97.5 %), while it was Enterobacter (47.5 %), and Pseudomonas (80 %) in 3HDJ and 4HDJ, respectively.The number of bacterial genera in 1CDJ, 2CDJ, 3CDJ, and 4CDJ was confirmed to be 2, 5, 4, and 5 genus, respectively, with Tetragenococcus being as the main genus in proportions of 65 %, 50 %, 45 %, and 42.5 %.Additionally, Staphylococcus (1CDJ: 35 %, 2CDJ: 37.5 %, and 3CDJ: 50 %) was found to be the major genus in CDJ, except in 4CDJ, in which is the main genus was in Bacillus (47.5 %).The bacterial population, in terms of the species found in HDJ and CDJ, is depicted in Fig. 2B.Seven species were found in the 1HDJ, with Bacillus subtilis predominating and accounting for 47.5 % of the species (19 clones).In reality, the 2HDJ contained 7 species, with Bacillus licheniformis predominating and accounting for 42.5 % of the species (17 strains).The Bacillus genus dominated the significant majority of the bacterial populations in samples from 1HDJ and 2HDJ (97.5 %), Clostridium and Staphylococcus accounting for remaining 2.5 %.Even though 11 species, including Alcaligenes sp., were found in 3HDJ, Enterobacter sp.predominated with a 25 % presence (10 clones).In the extremely varied genus, Enterobacter dominated with 47.5 % of the total genus population, followed by Halomonas and Chromohalobacter with 17.5 % each.Eight species were found in 4HDJ, and a large percentage of these (67.5 %) were Pseudomonas sp.species (27 clones).The Pseudomonas genus predominated in 4HDJ, accounting for 80 % the bacterial population.Additionally, Staphylococcus, Propionibacterium, Enterobacter, and Bacillus each contributed 2.5 % of the bacterial genus, and whereas Enterococcus accounted for 10 %.Only two species, Tetragenotoccus halophilus (26 clones/65 %) and Staphylococcus sp., were discovered in the 1CDJ.The 2CDJ bacterial analysis led to the identification of 7 species, with the T. halophilus species being predominantly present (42.5 %; 17 clones).In 3CDJ, Staphylococcus sp. was the dominant species, accounting for 47.5 % of all species (19 clones).T. halophilus (14 clones/35 %) and Tetragenococcus muriaticus (4 clones/ 10 %) also predominated in the bacterial populations in 3CDJ.Of the 13 species found, T. halophilus predominated in 4CDJ, accounting for 30 % of the strains (12 strains), exactly as it did in 1CDJ and 2CDJ.Bacillus sonorensis also predominated, accounting for 12.5 % of the clones (5 clones), while Alcaligenes faecalis, Alcaligenes sp., Lactiplantibacillus planrum, Leuconostoc sp., and Tetragenococcus sp. each contributed 2.5 % (1 clone).1HDJ to 4HDJ included a total of 53 different bacteria, and The microbial composition of doenjang differs, depending on environmental conditions, the raw material, and salinity (Ryu et al., 2021).It has been reported that the metabolites of doenjang vary depending on its microbial diversity (Zhang et al., 2022).The pH has been considered one of the most important indicators of microbial growth and metabolic features during fermentation (Chun et al., 2020;Zhang et al., 2022).During the doenjang fermentation process, pH decreases, which is caused by acid producing-bacteria.Among acid-producing bacteria, Staphylococcus, Tetragenococcus, and Weissella have shown relatively high contents in CDJ, suggesting that they play an important role in the fermentation process of CDJ.Additionally, these bacteria may have affected the pH and acidity of CDJ due to their production capacity of acid during the fermentation process (Song et al., 2021).In particular, Tetragenococcus, a halophilic lactic acid bacteria, was reported that dominated 12-15 % salt doenjang samples than other lower and higher salt samples (Chun et al., 2020).Therefore, Tetragenococcus is believed to be dominant in CDJ with a salinity of 12.80-14.60%.Bacillus, Enterococcus, Staphylococcus, and Tetragenococcus were dominant in doenjang samples (Fig. 2A), in accordance with previous results (Chun et al., 2020;Han et al., 2021;Lee et al., 2022).Chromohalobacter and Halomonas were identified only in 3HDJ and Pseudomonas was confirmed to be present only in 4HDJ (Fig. 2A).Chromohalobacter and Halomonas, which were halophilic bacteria, may have been derived from solar salt (Jung, Chun, & Jeon, 2015), and Pseudomonas in 4HDJ, might originated from meju (Kim, Han, Baek, Chun, & Jeon, 2022).In this way, the characteristics of the major bacteria in HDJ were confirmed to be different depending on the material used in manufacturing, whereas in CDJ, 2 to 3 genera of bacteria tended to be dominant.

Yeast diversity
The diversity of the yeast in HDJ and CDJ determined on the basis of 26S rRNA gene sequencing is summarized and presented in Fig. 2C and 2D, and supplementary Fig. S4 and S5.The yeast population in terms of the distribution of various genus in HDJ and CDJ is depicted in Fig. 2C.The number of yeast genera in 1HDJ, 2HDJ, 3HDJ, and 4HDJ was confirmed to be 10, 2, 4, and 6, respectively.Zygosaccharomyces was predominated in 1HDJ, 2HDJ, and 4HDJ in proportions of 20 %, 92.5 %, and 47.5 %, respectively, while the Candida genus accounted for 75 % the yeast population in 3HDJ.The number of yeast genera in 1CDJ, 2CDJ, 3CDJ, and 4CDJ was confirmed to be 7, 6, 5, and 4, respectively.Candida and Zygosaccharomyces were predominant in 1CDJ (32.5 % and 37.5 %) and 4CDJ (67.5 % and 27.5 %).Millerozyma, the prominent genus in 2CDJ and 3CDJ, occurred with a frequency of 47.5 % and 70 %, respectively.In terms of species, 13 species were identified in 1HDJ, with Zygosaccharomyces pseudorouxii dominating and accounting for 20 % (8 clones) of the species.Only two species, Millerozyma farinosa and Z. pseudorouxii, were recognized in 2HDJ, accounting for 7.5 % (3 clones) and 92.5 % (37 clones) of the species, respectively.In 3HDJ, 5 species of yeast were found; Candida versatilis representing 72.5 % (29 clones) of the clones, was the major species, while Yamadazyma triangularis accounted for 17.5 % (7 clones) of the species, and other yeast represented 10 % (4 clones) of the species.7 species were discovered in 4HDJ, including Z. pseudorouxii and Debaryomyces hansenii, accounting for 47.5 % (19 clones) and 30 % (12 clones) of the species, respectively.7 species were discovered in 4HDJ, including Z. pseudorouxii and D. hansenii, accounting for 47.5 % (19 clones) and 30 % (12 clones) of the species, respectively.Other yeasts accounted for less than 10 % of the species.Z. pseudorouxii was shown to be the most prevalent species in all the samples of HDJ, except for 3HDJ (Fig. 2D).In the case of 1CDJ, 10 species were identified, with Z. pseudorouxii being prominent and accounting for 37.5 % (15 clones), followed by Candida etchellsii was accounting for 20 % (8 clones) of the species.9 species were identified in 2CDJ; a relatively high proportion was Z. pseudorouxii (25 %; 10 clones) and M. farinosa (47.5 %; 19 clones).In 3CDJ, 5 species, including M. farinosa, were identified, and M. farinosa predominated accounting for 70 % (28 clones) of the species along with Ogataea polynirpha and Z. pseudorouxii which accounted for 10 % each.7 species were discovered in 4CDJ, and 45 % of them belonged to Candida mogii (18 clones), while Z. pseudorouxii accounted for 27.5 % (11 clones).CDJ analysis revealed that M. farinosa and Z. pseudorouxii frequently predominated in doenjang, but not in all samples of CDJ.A total of 21 yeast strains were found in 1HDJ to 4HDJ, and their phylogenetic relationship is shown in Fig. S4. C. versatilis and Z. pseudorouxii ATCC 42881 were the two most prevalent yeast strains among the 21 strains, with 30 and 60 clones of the respective strains being prevalent among the 160 clones of each species as shown in Fig. S4.Both C. versatilis and Z. pseudorouxii ATCC42981 displayed 96 to 99 % and 99 to 100 % similarity, respectively (not shown).Additionally, D. hansenii TJY37 and Y. triangularis NRRLY-5714 were confirmed to constitute 12 and 10 clones among the 160 clones, respectively (not shown), and displayed 99 % similarity.Fig. S5 depicts the phylogenetic between the 22 yeast strains discovered in the CDJ samples.M. farinosa CBS185 and Z. pseudorouxii ATCC42981 were predominant among the 22 strains, accounting for 40 and 23 clones, respectively, among the 160 clones identified (not shown).Next, C. mogii, Z. pseudorouxii ABT301, and C. etchellsii 33Z2 constituted 18, 17, and 13 clones, respectively, and all exhibited above 99 % similarity.
According to previous studies, Aspergillus, Debaryomyces, Millerozyma, Candida, and Zygosaccharomyces are the predominant fungi and yeast in doenjang, Candida and Zygosaccharomyces were mainly present in HDJ and CDJ (Chun et al., 2020;Han et al., 2021;Lee et al., 2022).Unusually, Aspergillus, known as the dominant fungus in meju and doenjang and used as a starter in commercial doenjang, was identified only in 1CDJ.Zygosaccharomyces is known as a genera that improves the flavor to soy foods (Ryu et al., 2021).Also, Candida is known to contribute to the production of aroma components (Tanaka, Watanabe, & Mogi, 2012).Therefore, the metabolites produced by Candida and Zygosaccharomyces during the fermentation process are likely to be the main substances contributing to the flavor of HDJ and CDJ.On the other hand, Debaryomyces was confirmed only in 4HDJ.It is judged that Debaryomyces is derived from the solar salt among the ingredients used at 4HDJ, in accordance with previous result (Chun et al., 2020).Another abundant genus, Millerozyma, has all species identified as M. farinosa.M. farinosa is commonly found in foods such as fermented liquor, miso, and soybean paste, and has been used in food production and fermentation processes (Mallet et al., 2012).Therefore, it appears to be dominant in CDJ, and in particular, 3CDJ was present at the highest rate of 70 %.In accordance with this study, Millerozyma, Candida, and Zygosaccharomyces are the predominant fungi and yeast associated with fermentation in CDJ.Mostly, but not all, HDJs, Candida and Zygosaccharomyces were shown as fermenting yeast.

Comparison of FAAs in the HDJ and CDJ samples
The results of the analysis of FAA contents in HDJ and CDJ samples, are shown in Fig. 3 and Table S2.There was no statistically significant difference between HDJ and CDJ in terms of overall FAAs content.Proline, aspartic acid, glutamic acid, alanine, leucine, phenylalanine, and lysine were detected as the major FAAs in HDJ and CDJ.The largest concentration of these was found to be glutamic acid.The major FAAs had no significant correlation with microorganisms (Fig. 3).The FAAs content of 1HDJ, 2HDJ, 3HDJ, and 4HDJ, respectively, was 5.37, 5.74, 2.92, and 2.93 mg/g.The FAAs concentration of 1CDJ, 2CDJ, 3CDJ, and 4CDJ was 6.78, 6.42, 0.73, and 7.22 mg/g, respectively.Among other essential amino acids, the concentration of leucine ranged from 1.48 to 2.73 mg/g in HDJ samples and from 1.07 to 3.83 mg/g in CDJ samples.Lysine showed a high concentration of 1.27 to 2.02 mg/g in HDJ samples and 0.98 to 3.00 mg/g in CDJ samples.Specifically, ammonia concentration was 0.78, 0.84, 0.49, and 0.48 mg/g in 1HDJ, 2HDJ, 3HDJ, and 4HDJ, respectively, and it was higher in HDJ than that in CDJ.Ammonia concentration had a positive correlation with Bacillus (r = 0.87, p = 0.05) (Fig. 3A).
Amino acids in doenjang are one of the most important metabolites because they contribute to the taste and flavor of doenjang and can be further converted into VFCs (Chun et al., 2020;Kim, Kwak, Kim, & Jeong, 2020).Glutamic acid is known as the primary amino acid in doenjang; this study confirmed the same, except for 3CDJ (Chun et al., 2020;Han, Chun, Kim, & Jeon, 2021).Glutamic acid is related to the unique umami taste and provides a savory taste to the doenjang.In addition, in our study, aspartic acid, which contributes to the umami taste, was confirmed to be in high concentration (Gao et al., 2021), followed by proline, alanine, leucine, lysine, and phenylalanine.Likewise, other amino acids, alanine and lysine, which provide a sweet taste, were present at high levels as confirmed in 1CDJ and 4CDJ.Proline, leucine, and isoleucine, which contribute to the bitter taste, were also the principal amino acids detected in doenjang (Namgung et al., 2010).
Tyrosine, ornithine, and lysine are the precursors of biogenic amines such as, tyramine, putrescine, and cadaverine.In the previous study, it was reported that tyramine production during the fermentation of ganjang might be caused by several bacteria, including Bacillus, Enterococcus, and Tetragenococcus (Kim, Chun, Kim, & Jeon, 2021).Further, Lactobacillus and Staphylococcus were associated with producing putrescine and cadaverine, respectively (Kim et al., 2021).1HDJ, 2HDJ, and 4CDJ have a comparatively high tyrosine content, and the distribution of Bacillus (97.5 %, 97.5 %, and 47.5 %) is quite high; in 4CDJ the Tetragenococcus (42.5 %) was genus predominated.Hence, the possibility of tyramine production seems high.In the previous report, most of the amino acids were decreased in most doenjang after reaching their highest level at the end of fermentation, resulting in a potential decrease in the taste (Chun et al., 2020;Han et al., 2021;Zhang et al., 2022).Therefore, 3HDJ and 4HDJ, appear to have decreased amino acids after the fermentation period, while 2CDJ and 3CDJ, do not attain the maximum amino acid contents at the beginning of fermentation.
Aldehydes are produced by lipid oxidation or decomposition products during fermentation, and these metabolic activities contribute to the fruity, nutty, and roasted flavors of soybean paste.Among them, 3methyl butanal provides a dark chocolatey and malty aroma (Jang, Yu, & Kim, 2021).It is a metabolite of leucine and is known as a major VFC in household doenjang, agreeing with previous studies, and has also been detected in CDJ samples (Lee, Kang, Kim, & Lim, 2015).3-methyl butanal had a positive correlation with Clostridium (p = 0.002), Cryptococcus (p = 0.002), Glomus (p = 0.002), Larrea (p = 0.002) (Fig. 4A), and Saccharomyces (p = 0.002) (Fig. 4B).The microorganisms associated with 3-methyl butanal included more yeast than bacterial, and these are believed to contribute to improving the flavor of doenjang.Furfural which provides a sweet, roasted, and woody aroma was detected in one HDJ (3HDJ) and three CDJ (2CDJ to 4CDJ) samples (Lee et al., 2015;Peng, Li, Shi, & Guo, 2014).In a previous study, it was reported to be the predominant aldehyde compound in household doenjang, but our results do not agree with the previous study.Meanwhile, Furfural had a positive correlation with Candida (p = 0.011), the main yeast of HDJ and CDJ (Fig. 4B).The Candida genus was one of the dominant yeasts in HDJ and CDJ, except 2HDJ.The Candida genus is known to encourage the synthesis of aromatic compounds.Among them, C. etchellsii and C. versatilis are halophilic late-maturing yeast used in soy sauce to produce of various aromatic compounds, might have contributed to enhancing of a mature fragrance to HDJ and CDJ (Tanaka et al., 2012).Lactiplantibacillus (p = 0.002), identified only in 3HDJ, also showed a positive correlation with Furfural, which can be used as a starter to improve the flavor of CDJ in the future (Fig. 4A).Isovaleric acid, a type of carboxylic acid, known for its unpleasant odor (dirty socks odor), which is generated during the long-term storage of doenjang, was detected only in HDJ samples and was found to be especially high in 4HDJ with a concentration of 12.39 % (Lee et al., 2015;Peng et al., 2014).Lee et al. (2015) also reported that the isovaleric acid component comprises a substantial percentage of VFCs in traditional doenjang, and this study also showed similar results.Since isovaleric acid can contribute to the unpleasant odor of HDJ and reduce its quality, it is judged to be a component that needs to be managed during the fermentation process.Isovaleric acid is produced by yeast protein metabolism (Peng et al., 2014).Among yeasts, Debaryomyces (p = 0.000) showed a positive correlation with isovaleric acid (Fig. 4B).In this study, not only yeast but also bacteria such as Pseudomonas, Propionibacterium, and Enterococcus also showed a positive correlation with isovaleric acid.For isovaleric acid component management, the occurrence of each microorganism should be prevented.Benzaldehyde, which has the simplest structure among aromatic aldehydes, has aromatic properties similar to almonds, has a woody odor, and is easily oxidized in air, was detected in all samples except for 4CDJ (Lee et al., 2015;Peng et al., 2014).2-pentyl furan is a metabolite found in or produced by Saccharomyces cerevisiae, Aspergillus, and bacteria (Xi et al., 2020).It is usually formed from the oxidation of linoleic acid and maillard reaction and delivers fruity, floral, buttery, green, beany, and roasted meat aromas to foods (Wang et al., 2021).2-phentyl furan, a kind of furan derivative, was detected in all HDJ sample, was detected exceptionally in 4CDJ.It was determined to result in increased linoleic acid oxidation and maillard reaction in HDJ, due to a relatively long fermentation period.In contrast, this reaction would not have occured in CDJ, which has a relatively short fermentation period.2phentyl furan had a positive correlation with Staphylococcus (p = 0.042) and had a negative correlation with Yamadazyma (p = 0.036) (Fig. 4B).Benzeneacetaldehyde, aldehyde type, was confirmed to be the major VFC in HDJ and CDJ.It is generated by precursor metabolites such as amino acids, hexose, and ribose (Watanabe et al., 2015).Especially, an increase in benzeneacetaldehyde was associated with increased phenylalanine levels and contribute to flowery odor (Tamura, Iwatoh, Miyaura, Asikin, & Kusano, 2022;Zhang et al., 2022).Benzeneacetaldehyde had a negative correlation with Staphylococcus (p = 0.042) and a positive correlation with Hanseniaspora (p = 0.040) (Fig. 4A).Nonanal is a VFC with an unpleasant odor (fatty and citrus odor) commonly detected in HDJ, and according to the results of this study, it was identified as one of the components to be managed in traditional soybean paste along with isovaleric acid.(Jiang, Lu, Ma, Liu, & He, 2023).One common lactic acid bacteria for soybean fermentation is T. halophilus, which may create VFCs, such as benzaldehyde, furfural, and methyl acetate, as well as organic acids (such as lactic and acetic acid), FAAs, and different VFCs, which might impact the flavor of final products.(Tanaka et al., 2012).Inoculation of T. halophilus can improve the flavor of protein fermentation products (Cui, Zheng, Wu, & Zhou, 2014).In this study, T. halophilus showed a high distribution only in the CDJ.It seems that the manufacturer of the CDJ also added T. halophilus to improve the flavor.While, Tetragenococcus had a negative correlation with diallyl disulphide (r = -0.91,p = 0.002) (Fig. 4A).Diallyl disulphide was detected in all samples, and HDJ had a larger amount than CDJ.It contribute to alliaceous, onion, garlic, and pungent odor (Zhou et al., 2023) and was a representative active substance in garlic (Asdaq et al., 2022).HDJ would be contaminated with pathogenic microorganisms due to spontaneous fermentation, so it seems that garlic or another material similar to garlic with antibacterial activity was used to make the HDJ.Previous studies reported that esters among VFCs were identified most frequently and were present in the highest amounts in commercial and traditional China doenjang (Zhang et al., 2021), miso (Japanese doenjang) (Inoue et al., 2016), and Korean traditional doenjang (Lee et al., 2015).But, in this study, the main VFCs in HDJ and CDJ samples were the aldehydes, including 3-methyl butanal, benzaldehyde, and benzeneacetaldehyde.Aldehydes may have contributed significantly to the flavor of HCD and CDJ because they were detected in high amounts and had a low odor threshold (Inoue et al.,2016).Clostridium, Cryptococcus, Glomus, Larrea, Saccharomyces, and Staphylococcus may have been involved in accordance with this study.

Comparison of isoflavone, TP, and TF contents in the HDJ and CDJ samples
Typical HPLC chromatograms of isoflavones in HDJ and CDJ samples, similar to the results shown in Fig. 5. Fig. 6A shows the changes in isoflavone glycoside and aglycone contents in the HDJ and CDJ samples.The glycoside concentrations in CDJ were 2 to 3 times higher than in HDJ.Especially, the 3HDJ never detected glycosides and just detected aglycones.Notably, the highest aglycone concentrations were found in the 1HDJ and 2HDJ samples, which were 1460.80 and 1314.22 µg/g, respectively.The 4CDJ sample showed the highest concentration of aglycones, approximately 1212.47 µg/g, and the lowest amounts of glycosides, approximately 150.47 µg/g, among the CDJ samples.Trace H. Yul Lee et al. amounts of acetylglycosides were detected in the CDJ samples, while none were found in the HDJ samples.In particular, the 3HDJ sample showed only aglycones (such as daidzein, glycitein, and genistein) with a complete absence of glycosides (including daidzin, glycitin, genistin, malonyldaidzin, malonylglycitin, malonylgenistin, acetyldaidzin, acetylglycitin, and acetylgenistin) (Fig. 6A).Moreover, the HDJ samples showed a greater enrichment of daidzein, glycitein, and genistein compared to the CDJ samples (Fig. 6A).And daidzein, glycitein, and genistein contents had a positive correlation with Bacillus, and a negative correlation with Staphylococcus except glycitein (Fig. 7A).The differences in TP and TF contents in the HDJ and CDJ samples are presented in Fig. 6B.In the case of 1HDJ and 2HDJ samples, the concentration of TP was 1.0 mg/g, and that of TF was 5.0 mg/g.In addition, TP was detected at a concentration 0.2 mg/g in both the 3HDJ and 4HDJ samples, but TF was found at 1.8 mg/g and the TF at 3.0 mg/g in the 3HDJ and 4HDJ samples.While, TF was recorded at approximately 2.0 mg/g for the 1CDJ, 2CDJ, and 3CDJ samples, it was remarkably higher, i.e., approximately 4.7 mg/g for the 4CDJ sample.Overall, the TF contents were greater in the HDJ samples compared to the CDJ samples.The amounts of TP contents were higher in the CDJ samples than those found in the HDJ samples.
Aglycones proportation were high in HDJ, and glycosides, malonyl glucoside.On the other hands, acetyl glucoside proportation were high in CDJ, in line with earlier data (Lee et al., 2023).The TF had a large amount compare to TP in all sample, in line with previous study (Lee et al., 2021).However, there was a difference from the Kwon et al. (2019) reports in which soybean fermentation food including doenjang had higher TP contents than TF contents.TP contents were high in CDJ than HDJ, unlike previous studies (Soung et al., 2021).The TF contents in 1HDJ and 2HDJ were significantly higher than those that of the CDJ samples.In general, the fermentation time of HDJ was about 1 year, while the fermentation time for CDJ was 180 days.The difference in TP and TF contents was interlinked with several factors, i.e., the processing time, ingredient types, and microbial diversity (Ryu et al., 2021).Therefore, this result was judged that because the plant cell walls' glycosides are degraded to form aglycone, phenol, and flavonoid compounds by hydrolytic enzymes like β-glucosidase, decarboxylase, and esterase or microbial acids that are produced by yeast and bacteria during fermentation (Park, Park, & Chang, 2019).It was reported that fermentation of soybean foods using Bifidobacterium longum, Lactiplantibacillus plantarum, and Levilactobacillus brevis corresponded to a decrease in the concentration of glycoside as the fermentation process    progressed (Hwang et al., 2021;Lee et al., 2018b;Lee et al., 2022).Since the fermentation period of CDJ was almost half that of the HDJ, glycoside bioconversion to aglycones was lower in CDJ, in line with the previous study, that glucoside, malonyl glucoside, and acetyl glucoside decreased and aglycone increased during the fermentation period (Kwak, Son, Chung, & Kown, 2015).In contrast, the concentrations of flavonoids were higher in CDJ (Fig. 6B).

Comparison of α-glucosidase and pancreatic lipase inhibitory
activities in the HDJ and CDJ samples α-glucosidase and pancreatic lipase inhibitory activities were compared between the HDJ and CDJ samples (Fig. 8).The α-glucosidase inhibitory activity was increased in response to the increased concentrations of doenjang extracts (0.25 to 1.0 mg/mL).The highest α-glucosidase inhibition was observed at a concentration of 64 to 78 % at 1.0 mg/mL of doenjang extracts (Fig. 8A).In contrast, trace levels of α-glucosidase inhibitory activity were observed in the CDJ samples 1CDJ, 2CDJ, and 3CDJ, which were about 10 to 13.5 fold lower than in the HDJ samples.However, the 4CDJ samples exhibited significantly higher α-glucosidase inhibitory activity than the other HDJ samples.
According to these findings, the CDJ samples had significantly lower α-glucosidase inhibitory activity than the HDJ samples.A similar observation was found for pancreatic lipase inhibitory activity.At a concentration of 1.0 mg/mL of doenjang extracts, the 1HDJ, 2HDJ, 3HDJ, and 4HDJ samples showed a maximum inhibitory activity of 56 to 68 %.In contrast, the 1CDJ, 2CDJ, and 3CDJ samples demonstrated pancreatic lipase inhibitory activity below 7 %, but the 4CDJ samples showed remarkably higher activity than the other CDJ samples.None of the CDJ sample extracts have shown pancreatic lipase inhibitory activities comparable to the HDJ samples at any concentration (Fig. 8B).
The α-glucosidase and pancreatic lipase inhibitory activities were high in HDJ than CDJ (Fig. 8).The α-glucosidase inhibitory activity was 38.9-78.2% in HDJ and 2.64-31.11% in CDJ at 1.0 mg/mL.These values were relatively high than by Shukla et al. (2016).Previous studies have shown a positive relationship between increased concentrations of isoflavone aglycones in soybean related fermented extracts and inhibition of α-glucosidase and pancreatic lipase (Lee et al., 2018b;Lee, Hwang, Son, & Cho, 2019;Lee et al., 2022).In this study, the contents of aglycones were higher in the HDJ extracts.Therefore, they considered that α-glucosidase and pancreatic amylase inhibitory activities were higher in the HDJ extracts (Fig. 8).

Conclusion
In this study, the differences in microbial diversity, FAA, VFC and isoflavone profiles and biological activities in HDJ and CDJ were analyzed.The major bacterial genus in 1HDJ and 2HDJ was Bacillus (97.5 %), while in 3HDJ and 4HDJ it was Enterobacter (47.5 %), and Pseudomonas (80 %), respectively.The main bacterial genus of CDJ was Tetragenococcus, occuring in proportions of 65 % (1CDJ), 50 % (2CDJ), 45 % (3CDJ), and 42.5 % (4CDJ).The Zygosaccharomyces genus among yeasts were found in all samples of HDJ and CDJ.The total FAA contents did not show a significant difference in HDJ and CDJ.But, glutamic acid, associated with a unique umami taste, predominated in HDJ and CDJ, together.The VFCs of 3-methyl butanal, benzeneacetaldehyde, and diallyl disulphide were detected in all samples of HDJ and CDJ.The highest concentration of 3-methyl butanal, which contributes to a dark chocolatey and malty aroma, was found in 1HDJ.The isoflavoneglycoside and TP contents in CDJ were 2 to 3 times higher than in HDJ.Notably, isoflavone-aglycone and TF contents in HDJ were higher than in CDJ, except for 3HDJ and 4CDJ.Our results provide, it is suspected that Bacillus is correlated with TPs, isoflavone-aglycones, Clostridium, Cryptococcus, Glomus, Larrea, and Saccharomyces are correlated with VFCs that improve the flavor of doenjang.However, no correlation has been observed with FAAs and TFs.The characteristics of microbes and components of doenjang according to the manufacturing methods, which can be used as basic data for the development of household and commercial doenjang in South Korea.In addition, microbes that show a correlation with the flavor of doenjang be used to new starters in the improvement of doenjang.However, studies with a larger amount of samples are necessary, to clarify the correlation between microbes and metabolites.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .
Fig. 1.Difference of relative values from the average values of physicochemical properties of household and commercial doenjang.

Fig. 2 .
Fig. 2. Comparisons of microbial population in household and commercial doenjang.(A) Bacterial population as genus, (B) Bacterial population as species, (C) Yeast population as genus, and (D) Yeast population as species.

Fig. 3 .
Fig. 3.The correlation heatmap between major free amino acid contents and microorganisms in household and commercial doenjang.(A) Bacterial composition as genus, (B) Yeast composition as genus.

Fig. 4 .
Fig. 4. The correlation heatmap between major volatile flavor compounds and microorganisms in household and commercial doenjang.(A) Bacterial composition as genus, (B) Yeast composition as genus.

Fig. 6 .
Fig. 6.Comparison of isoflavone, total phenolic, and total flavonoid contents in household and commercial doenjang.(A) Isoflavone content and (B) Total phenolic Total flavonoid contents.All values are presented as the mean ± standard deviation of triplicate determination.Different letters above the bars indicate significantly different (p < 0.05).

Fig. 7 .
Fig. 7.The correlation heatmap between isoflavone contents and microorganisms in household and commercial doenjang.(A) Bacterial composition as genus, (B) Yeast composition as genus.

Fig. 8 .
Fig. 8. Comparisons of digestive enzymatic inhibition activities in household and commercial doenjang.(A) α-Glucosidase inhibition activity and (B) Pancreatic lipase inhibition activity.All values are presented as the mean ± standard of triplicate determination.Different letters above the bars indicate significantly different (p < 0.05).