i An update to this article is included at the end

Food Research International 149 (2021) 110668

Contents lists available at ScienceDirect

Food Research International

journal homepage: www.elsevier.com/locate/foodres

Effects of the main ingredients of the fermented food, kimchi, on bacterial

composition and metabolite profile
Hye Seon Song a, 1, Se Hee Lee a, 1, Seung Woo Ahn a, Joon Yong Kim a, Jin-Kyu Rhee b, *, Seong
Woon Roh a, *
a
Microbiology and Functionality Research Group, World Institute of Kimchi, Gwangju 61755, Republic of Korea
b
Department of Food Science and Engineering, Ewha Womans University, Seoul 03760, Republic of Korea

A R T I C L E I N F O A B S T R A C T

Keywords: Kimchi is a fermented food prepared via spontaneous fermentation by lactic acid bacteria originating from raw
Fermented food ingredients. To investigate the effect of these ingredients on food fermentation, four types of food that differed
Ingredient only in their main raw ingredients (kimchi cabbage, green onion, leaf mustard, and young radish) were evalu­
Lactic acid bacteria
ated. The major microorganisms were Leuconostoc gelidum, Weissella kandleri, and Lactobacillus sakei groups. The
Metataxonomics
distribution of these species depended on the sample type. All three species were primarily distributed in the food
Metabonomics
prepared from kimchi cabbage and young radish; however, the Lac. sakei group was hardly found in the food
prepared using green onion and leaf mustard. Metabolite analysis results showed that the free sugar, organic
acid, ethanol, and amino acid profiles differed with the sample type. This study indicates that the main in­
gredients could be an important factor in determining the composition of the microbial community and the
metabolite composition.

1. Introduction and antimicrobial compounds, through fermentation (Lee et al., 2020).

Therefore, the sensory characteristics of kimchi vary depending on the
Kimchi is a traditional fermented food in Korea, and there are several dominant LAB type during fermentation. To manufacture high-quality
types depending on the available seasonal ingredients and the fermented food with superior taste, LAB with excellent fermentation
manufacturing methods used (Lee, Whon, Roh & Jeon, 2020). Kimchi ability and functionality are used as starter.
cabbage [Brassica rapa L. var. pekinensis (Lour.)], radish (Raphanus sat­ LAB are typically found in the main ingredients of kimchi, such as
ivus L.), green onion (Allium wakegi Araki), leaf mustard [Brassica jun­ kimchi cabbage and garlic, and play an important role in determining
cea (L.) Czern], and young radish (Raphanus sativus L.) are typically used kimchi fermentation (Lee, Jung & Jeon, 2015; Song et al., 2020). To
as the main ingredients for the various preparations of this food, and it is date, several investigations have been carried out to determine the ef­
fermented after seasoning with other ingredients, such as garlic, ginger, fects of temperature, seasoning, salt concentration, presence of salt-
salt, red pepper, and salt-fermented fish. fermented fish, and LAB starter type on kimchi fermentation. Food
Kimchi fermentation is influenced by the ingredients, fermentation components and external factors are the main determinants of the mi­
temperature, salt concentration, oxygen availability, and pH, which crobial community. There are currently more than 100 types of kimchi
determine the taste and quality of the final fermented product (Jung, Lee (Ji et al., 2013). The previous studies primarily focused on the repre­
& Jeon, 2014). Due to the activities of various lactic acid bacteria (LAB), sentative kimchi type, prepared using kimchi cabbage, and there are few
predominantly Leuconostoc, Lactobacillus, Weissella, Pediococcus, and studies on kimchi prepared using other ingredients. Therefore, this study
Lactococcus (Kim, Bang, Beuchat, Kim & Ryu, 2012), during the determined the effects of the intrinsic microorganisms present in the
fermentation process, kimchi may be unevenly fermented each time, and different main ingredients of this food on its fermentation. To investi­
fermentation characteristics may vary. LAB produce various com­ gate the food fermentation characteristics, microbial community struc­
pounds, such as lactate, mannitol, ethanol, vitamins, carbon dioxide, ture, and metabolite production in food samples prepared with different

* Corresponding authors.
E-mail addresses: jkrhee@ewha.ac.kr (J.-K. Rhee), swroh@wikim.re.kr (S.W. Roh).
1
These authors contributed equally to this work.

https://doi.org/10.1016/j.foodres.2021.110668
Received 31 March 2021; Received in revised form 5 August 2021; Accepted 24 August 2021
Available online 28 August 2021
0963-9969/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
H.S. Song et al. Food Research International 149 (2021) 110668

main ingredients, but with the same seasonings, the samples were centrifugation (18,000g, 10 min) of 1 mL of each sample collected at
evaluated using high-throughput sequencing and proton nuclear mag­ different times using a FastDNA SPIN Kit for soil (MP Biomedicals, Santa
netic resonance (1H NMR) spectroscopy. Ana, CA, USA), according to the manufacturer’s instructions. The con­
centrations of the extracted DNA were determined using a spectropho­
2. Materials and methods tometer (Nanodrop Technologies, Wilmington, DE, USA), and the
quantity and purity were estimated via 1% agarose gel electrophoresis.
2.1. Preparation of kimchi samples Microbial succession in each sample was monitored by amplifying the
V3–V4 region of the 16S rRNA gene sequence using the bacterial primers
Four types of kimchi were prepared using four different main in­ 341F and 805R containing adapter sequences, as previously described
gredients: kimchi cabbage, green onion, leaf mustard, and young radish (Fadrosh et al., 2014; Song et al., 2020). Three amplified PCR products
(Fig. 1). The kimchi was prepared by mixing ingredients in the following were visualized using 1% agarose gel electrophoresis and purified using
proportions: 85% main ingredient, 2.5% garlic, 3% red pepper, 1.8% the QIAquick PCR purification kit (Qiagen, Hilden, NRW, Germany).
salt, and 7.7% tap water. The four samples (sample C prepared using Secondary PCR amplification was performed using i5 and i7 index
kimchi cabbage; sample G prepared using green onion; sample L pre­ primers to attach the Illumina Nextera barcodes to both amplicon ends.
pared using leaf mustard; sample Y prepared using young radish) were The indexed PCR products were quantified using the Quant-iT Pico­
stored at 4 ◦ C for 30 days. During the fermentation period, pH and LAB Green dsDNA Assay Kit (Invitrogen, Carlsbad, CA, USA). Three repli­
counts were estimated, and microbial community diversity and metab­ cates per sample were pooled at equimolar concentrations. Amplicons
olite changes were analyzed, as described below. were paired-end sequenced using a MiSeq sequencing platform (Illu­
mina Inc., Albany, NY, USA) with V2 chemistry following the manu­
2.2. Sampling and measurement of lactic acid bacterial viability and pH facturer’s instruction.

The soup of the food sample was periodically collected from the 2.4. 16S rRNA gene sequencing data processing
samples at 0, 5, 15, and 30 days. pH values were measured using a pH
meter (Thermo Fisher Scientific, Waltham, MA, USA) in triplicates. One Paired-end sequencing data were filtered using Trimomatic ver. 0.32
mL of the soup was used for the enumerations of LAB. Large particles in to check the read quality, and low-quality reads with Phred quality score
the soup were removed using a sterile stomacher filter bag (Nasco Whirl- below 25 were removed (Bolger, Lohse & Usadel, 2014). The filtered
Pak, Fort Atkinson, WI, USA). The filtered soups were serially diluted paired-end reads were merged using PANDAseq (Masella, Bartram,
with sterilized 0.85% saline solution and inoculated onto 3M™ Petrifilm Truszkowski, Brown & Neufeld, 2012), and the primer sequences were
LAB count plates (3M™ Microbiology, Saint Paul, MN, USA) in tripli­ removed. Sequences were denoised using DUDE-Seq. Taxonomic as­
cates. The plates were incubated at 30 ◦ C for 2 days, and the microbial signments were prepared using the USEARCH program with EzBioCloud
counts were expressed as CFU/mL of the soup. database, and the similarity was calculated through pairwise alignment.
Potential chimeric sequences were identified and removed using
2.3. DNA extraction and 16S rRNA gene-amplicon sequencing UCHIME (Edgar, Haas, Clemente, Quince & Knight, 2011). The
remaining sequences with <97% similarity were clustered into opera­
Total DNA was extracted from the pellets obtained through the tional taxonomic units (OTUs) using CD-HIT and UCLUST (Edgar,

Fig. 1. Summary of each type of kimchi.

2
H.S. Song et al. Food Research International 149 (2021) 110668

2010). Microbial diversity in each sample was estimated based on differences among the sample types.
Chao1, abundance-based coverage estimators (ACE), Shannon index,
and Simpson’s index (Chao, 1987; Chazdon, Colwell, Denslow & Guar­ 3. Results
iguata, 1998; Hill, 1973; Shannon, 1997).
3.1. Changes in fermentation properties based on the specific sample
2.5. Metabolome analysis
The changes in the pH and LAB counts of the four types of kimchi
Metabolites were analyzed in three replicates using 1H NMR spec­ (samples C, G, L, and Y) during food fermentation were determined and
troscopy. Supernatants obtained through the centrifugation (18,000g, are presented in Fig. 2A and B. At the beginning of fermentation, the
10 min) of 500 μL of each sample were mixed with 500 μL of deuterium average pH of all samples ranged from 4.44 to 4.68, which increased to
oxide supplemented with 10 mM 2,2-dimethyl-2-silapentane-5-sulfo­ 5.12–5.50 after 5 days. As fermentation progressed, the pH dropped in
nate sodium salt (DSS; final concentration of 5 mM) as an internal all samples, ranging from 3.9 to 4.36 after 30 days. Among the four
standard. The 1H NMR spectral analysis and identification of metabo­ samples, the pH of sample C decreased most quickly; however, the
lites from the samples were performed, as described previously (Song overall pH change followed the typical kimchi fermentation pattern,
et al., 2020). Briefly, 1H-NMR measurements were performed using a regardless of the sample type. Average LAB counts for the samples
Varian Inova 600-MHz NMR spectrometer (Varian Inc., Palo Alto, CA, initially ranged from 4.63 to 4.87 log CFU/mL. This increased to
USA). The 1H NMR spectra were corrected using Chenomx NMR Suite 5.46–5.75 log CFU/mL by day 5 of fermentation in samples C, G, and L,
version 8.3 (Chenomx, Edmonton, Alberta, Canada), using the internal whereas it rapidly increased to 6.99 log CFU/mL in sample Y. On day 15,
standard DSS. Automatic baseline correction and automatic phase the average LAB count in samples C, L, and Y ranged from 8.73 to 8.91
correction were performed according to the chemical form indicator. log CFU/mL, and in sample G, the count was 7.47 log CFU/mL, which
The quantification of the metabolites was performed using the built-in was less than that in other samples.
database of the Chenomx NMR Suite 8.3.
3.2. Alpha diversities of the bacterial community during food
2.6. Availability of data and materials fermentation

The 16S rRNA gene amplicon datasets presented in this study were To define the microbial community of the food samples prepared
deposited in the National Center for Biotechnology Information (NCBI), using different main ingredients, bacterial composition in the samples
the European Bioinformatics Institute (EBI), and the DNA Data Bank of was characterized based on 16S rRNA gene sequences. A total of
Japan (DDBJ) under the accession number PRJNA674725. 576,019 reads were obtained through the Illumina Miseq platform from
all samples, and the Good’s coverage values for each sample were
2.7. Statistical analysis 97.9–99.9%, based on clustering into OTUs at a 97% similarity level.
This suggested that the reads obtained from each sample were sufficient
Principal component analysis (PCA) was carried out using the for bacterial diversity analysis. Phylotype diversity of the bacterial
prcomp, devtools, and ggbiplot packages of the R program. To compare community in each sample was expressed via Chao1, ACE, Shannon, and
the difference in metabolite production between the samples during the Simpson indices (Table S1). The Chao1, ACE, and Shannon indices of the
fermentation period, statistical analysis was performed using Metab­ bacterial community composition in all samples decreased as the
oAnalyst 4.0. The metabolites that differed significantly among the fermentation started and stabilized after 15 days. In particular, sample Y
samples were analyzed using one-way ANOVA and Tukey’s HSD post- exhibited the highest Chao1, ACE, and Shannon indices of all food types
hoc test (p ≤ 0.001). Additionally, to determine the association be­ at day 0 of fermentation, and this decreased as fermentation progressed.
tween the microbial community and metabolites of each sample, a non-
similarity test of nonmetric multidimensional scaling (NMDS) was per­ 3.3. Bacterial community composition of the four sample types
formed using the Bray–Curtis pseudo-distance matrix in R with the
metaMDS and envfit (permutations = 999, p < 0.5) functions. Prism To compare the bacterial profile of the samples based on their main
software version 8.0 (GraphPad, La Jolla, CA, USA) was used to verify ingredients during the fermentation period, the V3–V4 region of the 16S
the Spearman correlation between major bacterial taxa and metabolites. rRNA gene was targeted in the DNA extracts from the samples. The
The analysis of similarities (ANOSIM) was used to determine the taxonomic compositions of the four types of food are shown in Fig. 3A, B,

Fig. 2. Variation in pH (A) and lactic acid bacteria counts (B) in each food type during the fermentation period. C, Food prepared using kimchi cabbage; G, green
onion; L, leaf mustard; and Y, young radish. The LAB count is expressed in log CFU/mL.

3
H.S. Song et al. Food Research International 149 (2021) 110668

Fig. 3. Relative abundance of total bacteria at the family (A) and genus (B) level, and lactic acid bacteria at the species level (C); PCA plot displaying the bacterial
communities in each type of the food (D) during fermentation. The bacterial composition was classified according to 16S rRNA gene sequence annotation. The x-axis
represents the sampling day for each sample, and the y-axis represents the relative abundance of bacteria (A–C). C, Food prepared using kimchi cabbage; G, green
onion; L, leaf mustard; and Y, young radish. The kimchi samples were collected at 0, 5, 15, and 30 days.

and C. During the initial phase of fermentation, the composition of the increased in samples C and Y during the late fermentation period. At the
bacterial community was varied in samples C, G, and Y, whereas in genus level, Weissella, Leuconostoc, and Lactobacillus primarily domi­
sample L, Pseudomonadaceae comprised 85.2% of the bacterial popula­ nated samples C and Y, whereas Leuconostoc and Weissella gradually
tion, thereby forming a simpler microbial community structure dominated samples G and L, respectively (Fig. 3B). Specifically, the
(Fig. 3A). Initially, the LAB were present in a small proportion; however, relative abundance of Weissella members within the total bacterial
from day 5 of fermentation, the levels of Leuconostocaceae increased population of sample L increased from <1% to >88.8%. At the species
rapidly in all types of kimchi, and Lactobacillaceae levels gradually level of the LAB, the Leuconostoc (Leu.) gelidum, Weissella (Wei.) kandleri,

4
H.S. Song et al. Food Research International 149 (2021) 110668

and Lactobacillus (Lac.) sakei groups were differentially predominant in between the samples, a heat map and PCA plot were constructed,
the mid- and late fermentation stages (5–30 days) of samples C and Y showing the changes in major metabolite contents against the fermen­
(Fig. 3C). However, sample G was primarily occupied by the Leu. gelidum tation periods (Fig. 5). The heat map constructed using the concentra­
group, and the Wei. kandleri (koreensis) group was predominantly tions of 39 metabolites showed the comprehensive difference between
observed in sample L from day 5 post-fermentation and then remained each food type during fermentation. Most primary metabolites were
stable. PCA was used to compare the difference between the four sam­ distributed similarly in all samples at the beginning of fermentation;
ples based on their microbial taxonomic profiles at the genus level however, as the fermentation progressed, the changes in the primary
(Fig. 3D). In the PCA plot, bacterial community profiles of each food and secondary metabolite levels differed for each sample (Fig. 5A).
sample differed, depending on the main ingredients, as fermentation Based on the metabolite profile at the end of fermentation (days 15 and
progressed (with PC1 and PC2 accounting for a total of 46.8% and 30), samples C, Y, and L were grouped into one cluster, whereas sample
29.0% of the variables, respectively). G was separated, because it exhibited different trends. In addition,
various metabolites, such as several organic acids, amino acids, and
vitamins, were produced at a relatively high concentration in sample Y
3.4. Metabolite changes in four sample types than that in the other samples. In the PCA plot, samples C, Y, and L
tended to move to the far right, in the opposite direction to the PC1
We investigated whether differences in the main ingredients of the dimension, as fermentation progressed. The metabolic profile of sample
food could affect the composition of the main metabolites during G progressed lesser than that of the others (Fig. 5B). ANOSIM results
fermentation (Supplementary Fig. S1). Metabolite analysis results showed no significant differences between the metabolite profiles of
showed that each sample type contained various components, such as samples C and L; however, there were significant differences between
free sugar, organic acids, ethanol, amino acids, and vitamins (Figs. 4 and the profiles of the other sample types (ANOSIM p ≤ 0.05). In addition,
5A, and Supplementary Fig. S2). The major free sugars present in each one-way ANOVA and post-hoc tests were performed based on the
sample were glucose, fructose, and sucrose at the start of fermentation. metabolite content of each sample (Supplementary Fig. S3). After 30
As the fermentation progressed, the changing patterns in the concen­ days of fermentation, 12 metabolites exhibited significant differences (p
trations of glucose, fructose, sucrose, ethanol, glycerol, and glycine were < 0.001) in the metabolite composition analysis in all samples.
similar in samples C and Y; however, mannitol concentration was rela­
tively higher in sample C, compared to that in sample Y. In sample G, the
concentrations of fructose, sucrose, and glycine decreased, whereas that 3.5. Relationships between bacterial community and metabolite profiles
of mannitol, butyrate, and ethanol increased, with the concentrations of
butyrate and ethanol being higher in sample G, compared to that in The NMDS plot was generated to determine the association between
other samples. The changes in major metabolites during fermentation in the bacterial community composition and the metabolites for each
sample L included a decrease in fructose and sucrose concentrations and sample type (Fig. 6A). The NMDS plot showed that the metabolite
an increase in mannitol, lactate, butyrate, and ethanol levels. Addi­ profiles correlated with 11 taxa of the bacterial populations (envfit, p <
tionally, to understand the differences in the composition of metabolites 0.5). The stress value of the NMDS plot was 0.05. As fermentation

Fig. 4. Changes in the composition of metabolites in the food prepared using kimchi cabbage (C), green onion (G), leaf mustard (L), and young radish (Y) during the
fermentation period.

5
H.S. Song et al. Food Research International 149 (2021) 110668

Fig. 5. (A) Heatmap of metabolite profiles in each type based on the fermentation period. (B) PCA plot of metabolite profiles in each food type. C, Food prepared
using kimchi cabbage; G, green onion; L, leaf mustard; and Y, young radish. The kimchi samples were collected at 0, 5, 15, and 30 days.

progressed, the four sample types showed a fermentation pattern that correlated with the production of ethanol, mannitol, and butyrate. The
moved to the right of NMDS1, involving the Leu. gelidum, Lac. sakei, and presence of Lac. sakei group negatively correlated with the utilization
Wei. kandleri groups. The correlation between the top 11 taxa and the glucose, fructose, and sucrose and positively correlated with the pro­
metabolite profiles of the samples was confirmed through a heatmap duction of ethanol, butyrate, and lactate. The presence of Wei. kandleri
(Fig. 6B). The presence of the Leu. gelidum group negatively correlated group was correlated with the production of lactate, ethanol, and
with the utilization of fructose and sucrose as free sugars and positively butyrate by utilizing fructose and sucrose, rather than glucose.

6
H.S. Song et al. Food Research International 149 (2021) 110668

Fig. 6. Analysis of associations between bacterial composition and metabolite profile. (A) Nonmetric multidimensional scaling (NMDS) plot of the associations
between bacterial composition and metabolite profile, conducted based on Bray–Curtis dissimilarities. The arrows indicate bacterial taxa variables with a significance
factor of p < 0.5. (B) Correlation between the top 11 bacterial taxa and the metabolite composition during food fermentation. Heat map was generated using
Spearman coefficients between the top 11 bacterial taxa and metabolite composition. C, Food prepared using kimchi cabbage; G, green onion; L, leaf mustard; and Y,
young radish. The food samples were collected at 0, 5, 15, and 30 days.

4. Discussion production.
During fermentation, there were slight differences in the pH and LAB
The fermentation of kimchi is greatly influenced by external factors, count between the samples. pH decreased faster in sample C, and LAB
such as temperature, salt concentration, and the ingredients and increased slower in sample G. In previous studies, the initial pH of the
manufacturing methods. The food type is determined by the main in­ food was confirmed to range from 5.0 to 5.5 (Jeong, Lee, Jung, Choi &
gredients used, and the manufacturing method and minor ingredients Jeon, 2013; Lee et al., 2015), whereas that in our study was initially
are applied accordingly. The food fermentation is initiated by various approximately 4.5, and then it subsequently decreased after increasing
microorganisms present in the major and minor ingredients; and to approximately 5.1–5.5. Sample G showed an increase in the pro­
therefore, main ingredients are the major factors that determine the duction of butyrate, compared to that of lactate and acetate (Fig. 4), and
microbial community and metabolite composition (Cheigh & Park, the pH in sample G did not reduce like that in sample C. In contrast, in
1994; Song et al., 2020). Metabolite production and food quality depend sample C, the pH was relatively lower than that in the other samples,
on the type of LAB species present in the main ingredients of the food. probably due to a difference in the amount of amino acids produced,
Therefore, our study investigated whether differences in the main in­ despite the low organic acid content (Fig. 4 and Supplementary Fig. S2).
gredients cause changes in the bacterial composition and metabolite The slow change in pH during late fermentation is due to the buffering

7
H.S. Song et al. Food Research International 149 (2021) 110668

action of free amino acids and inorganic ions (Kang, Kim & Byun, 1988). fermentation process of each food. In samples C and Y, Lac. sakei counts
Sample C was sensitive to the changes in pH, possibly due to low amino increased, and the glycerol levels decreased as the fermentation pro­
acids production, inducing a relatively low pH value, compared to that gressed (Fig. 3C and Fig. 4). In most Lac. sakei strains, the gene encoding
of the other samples. the enzyme (glycerol kinase; EC2.7.1.30) that converts glycerol to sn-
pH and LAB count are key indicators of the degree of food fermen­ glycerol 3-phosphate has been identified; however, further research is
tation. During fermentation, the pH is reduced by the acids, such as needed to determine whether Lac. sakei degrades glycerol (McLeod
organic and amino acids, produced by the LAB. Therefore, pH is influ­ et al., 2010). These results suggest that Lac. sakei can use glycerol as
enced by the composition of the microbial communities in the food. sugar for growth. Interestingly, when the Lac. sakei group dominated in
Consequently, we studied the changes in LAB count, with respect to the a food sample, lactate production of that sample was higher, compared
fermentation period. We observed no difference in the LAB count be­ to that of the other samples, due to their homolactic fermentation; and
tween the samples on day 0; however, as fermentation progressed, the therefore, the production of ethanol and mannitol was relatively low.
LAB count in sample Y increased most rapidly, and all samples reached The Leu. gelidum group was the predominant LAB group in sample G,
≥1 × 108 CFU/mL by the end of the fermentation period. These results and it appeared to be involved in mannitol, ethanol, and butyrate pro­
suggest that a relatively large LAB population grew rapidly in the rad­ duction through heterolactic fermentation using fructose and sucrose; it
ishes in sample Y. Bacterial abundance testing results suggest that the barely influenced lactate or glycerol production. Leu. gelidum is well
main ingredients influence the diversity of the bacterial community in known for mannitol production (Chun et al., 2017; Jung et al., 2014).
food. Therefore, the high microbial diversity of sample Y reflects the Furthermore, the heatmap showed no correlation between the glycerol
diversity of the bacteria that exist in young radish, the main ingredient content and the presence of the Leu. gelidum group in sample G. The PCA
of sample Y. There was an excessive growth of LAB in all samples; and and NMDS plots showed that the metabolite profiles of sample G at the
therefore, the bacterial richness in the food decreased as the fermenta­ beginning and end of food fermentation were closely related, such that
tion proceeded. The development of a highly acidic environment and their fermentation seemed to proceed more slowly than that in the other
low temperature hindered bacterial adaptation, resulting in a change in samples. In sample L, where the Wei. kandleri group was dominant,
the composition of the microbial community. ethanol, mannitol, and lactate were produced using fructose and sucrose
At the beginning of fermentation, various bacteria comprised the rather than glucose, indicating that the preferred free sugars differed
community of each sample type; however, as fermentation progressed, depending on the predominant LAB group, resulting in the production of
the Leu. gelidum, Lac. sakei (recently reclassified as Latilactobacillus sakei different metabolites.
(Zheng et al., 2020)), and Wei. kandleri groups were the most abundant The one-way ANOVA results showed that the levels of 12 of the 39
phylotypes in each food type. During the fermentation process, the metabolites, identified at 30 days of fermentation, differed significantly
predominant LAB species were similar to those previously reported (Cho between the groups (Supplementary Fig. S3A). Among them, the glucose
et al., 2006; Jung et al., 2014; Jung et al., 2011); however, the species content in sample Y decreased faster than that in other samples (Fig. 4),
and distribution of the predominant LAB differed depending on the food probably because it was consumed as the major carbohydrate by the Lac.
type. The microbial communities are affected by salt concentration, sakei group. The distribution of the Lac. sakei group increased as the
main ingredient used, fermentation period, sampling time, and nutritive fermentation progressed in sample Y, and it exhibited a negative cor­
components involved (Lee et al., 2015; Lee, Song, Jung, Lee & Chang, relation with the glucose content (Fig. 6B). In contrast, proline con­
2017; Lee, Song, Park & Chang, 2019). Red pepper closely affects the centration increased in sample L as fermentation progressed, which was
ontogeny of Wei. cibaria during food fermentation (Kang, Cho & Park, presumably due to the metabolism of the Wei. kandleri group. In addi­
2016). Therefore, the differences in the composition of the bacterial tion, the content of amino acids essential for human metabolism tended
community food could be attributed to the differences in the intrinsic to increase in samples L and Y, compared with that in the other samples,
microorganisms present in the main ingredients (Lee et al., 2015; Song as the fermentation progressed. Essential amino acids in fermented
et al., 2020). In this study, during the early fermentation stage, the foods can have direct or indirect effects on improving health (Steink­
different food samples exhibited significant differences in the composi­ raus, 1997). The increase in free amino acid content in samples L and Y
tion of the microbial community at the genus level (Fig. 3B). The genus could be attributed to the extracellular proteolytic activity of the LAB
Lactobacillus was more abundant in sample Y than that in sample C, present in the samples (Toe et al., 2019). The metabolite content of the
although the sample groups C and Y harbored similar major bacterial other samples exhibited relative quantitative differences corresponding
community members at the genus level as the fermentation proceeded. with the difference between their main ingredients. Therefore, our re­
The genus Leuconostoc was the most abundant in sample G, whereas sults suggest that kimchi of varying qualities can be prepared using
Weissella was dominant in sample L. These results indicate that the different main ingredients.
dominant bacterial phylotypes differ with the main ingredient. Among
the various endogenous microorganisms in the main ingredients, LAB 5. Conclusions
lead the fermentation and are the dominant microorganisms (Chang,
Shim, Cha & Chee, 2010). In this study, we investigated the effects of the main ingredients of
Various LAB are involved in the kimchi fermentation process, and kimchi on the composition of the bacterial community and metabolite
many metabolites are produced through fermentation, which de­ concentration during food fermentation. The composition of the bacte­
termines the taste and quality of the food (Cho et al., 2006; Chun, Kim, rial community and their distribution differed between food prepared
Jeon, Lee & Jeon, 2017; Moon, Kim & Chang, 2018). The metabolic using kimchi cabbage, green onion, leaf mustard, and young radish.
profiles of each type differed depending on the predominant LAB group, These differences in bacterial composition led to differences in the me­
such as the Leu. gelidum, Lac. sakei, and Wei. kandleri members, during tabolites produced in the food. Therefore, the main ingredients were a
fermentation. Fructose and sucrose were considered the major free major contributing factor to food quality. To improve the basic under­
sugars used in samples C, G, and L, whereas glucose, fructose, and su­ standing of kimchi fermentation, further research is needed to investi­
crose are all used as major free sugars in sample Y (Fig. 4), likely due to gate the influence of minor ingredients on food fermentation.
the difference in microbial community profiles between the food sam­
ples. The PCA and NMDS plots show the separation of each sample type CRediT authorship contribution statement
based on the metabolites produced (Fig. 5B and Fig. 6A). In the NMDS
plot, fermentation patterns were formed by the Leu. gelidum, Lac. sakei, Hye Seon Song: Investigation, Formal analysis, Visualization,
and Wei. kandleri groups according to the sample type, and the three Writing - original draft. Se Hee Lee: Investigation, Formal analysis,
groups were identified as important microorganisms influencing the Visualization, Writing - review & editing. Seung Woo Ahn:

8
H.S. Song et al. Food Research International 149 (2021) 110668

Investigation, Methodology. Joon Yong Kim: Formal analysis, Visuali­ sequencing on the Illumina MiSeq platform. Microbiome, 2(1), 6. https://doi.org/
10.1186/2049-2618-2-6.
zation. Jin-Kyu Rhee: Supervision, Writing - review & editing. Seong
Hill, M. O. (1973). Diversity and evenness: A unifying notation and its consequences.
Woon Roh: Project administration, Supervision, Conceptualization, Ecology, 54(2), 427–432.
Writing - review & editing. Jeong, S. H., Lee, S. H., Jung, J. Y., Choi, E. J., & Jeon, C. O. (2013). Microbial succession
and metabolite changes during long-term storage of Kimchi. Journal of Food Science,
78(5), M763–M769.
Declaration of Competing Interest Ji, Y., Kim, H., Park, H., Lee, J., Lee, H., Shin, H., … Holzapfel, W. H. (2013).
Functionality and safety of lactic bacterial strains from Korean kimchi. Food Control,
The authors declare that they have no known competing financial 31(2), 467–473.
Jung, J. Y., Lee, S. H., & Jeon, C. O. (2014). Kimchi microflora: History, current status,
interests or personal relationships that could have appeared to influence and perspectives for industrial kimchi production. Applied Microbiology and
the work reported in this paper. Biotechnology, 98(6), 2385–2393.
Jung, J. Y., Lee, S. H., Kim, J. M., Park, M. S., Bae, J.-W., Hahn, Y., … Jeon, C. O. (2011).
Metagenomic analysis of kimchi, a traditional Korean fermented food. Applied and
Acknowledgements Environmental Microbiology, 77(7), 2264–2274.
Kang, B. K., Cho, M. S., & Park, D. S. (2016). Red pepper powder is a crucial factor that
This research was supported by the World Institute of Kimchi funded influences the ontogeny of Weissella cibaria during kimchi fermentation. Scientific
Reports, 6(1), 1–8.
by the Ministry of Science and ICT (KE2101-2), Basic Science Research Kang, S.-S., Kim, J.-M., & Byun, M.-W. (1988). Preservation of Kimchi by ionizing
Program of the National Research Foundation of Korea (NRF) radiation. Journal of Food Hygiene and Safety, 3(2), 225–232.
(2018R1D1A1A09082921 and 2018R1D1A1B07045349), and Ewha Kim, J., Bang, J., Beuchat, L. R., Kim, H., & Ryu, J.-H. (2012). Controlled fermentation of
kimchi using naturally occurring antimicrobial agents. Food Microbiology, 32(1),
Womans University Research Grant of 2021, and High Value-added
20–31.
Food Technology Development Program, the Ministry of Agriculture, Lee, S. H., Jung, J. Y., & Jeon, C. O. (2015). Source tracking and succession of Kimchi
Food and Rural Affairs (MAFRA), Republic of Korea. lactic acid bacteria during fermentation. Journal of Food Science, 80(8),
M1871–M1877.
Lee, M., Song, J. H., Jung, M. Y., Lee, S. H., & Chang, J. Y. (2017). Large-scale targeted
Appendix A. Supplementary material metagenomics analysis of bacterial ecological changes in 88 kimchi samples during
fermentation. Food Microbiology, 66, 173–183.
Supplementary data to this article can be found online at https://doi. Lee, M., Song, J. H., Park, J. M., & Chang, J. Y. (2019). Bacterial diversity in Korean
temple kimchi fermentation. Food Research International, 126, 108592. https://doi.
org/10.1016/j.foodres.2021.110668. org/10.1016/j.foodres.2019.108592.
Lee, S. H., Whon, T. W., Roh, S. W., & Jeon, C. O. (2020). Unraveling microbial
References fermentation features in kimchi: From classical to meta-omics approaches. Applied
Microbiology and Biotechnology, 104(18), 7731–7744.
Masella, A. P., Bartram, A. K., Truszkowski, J. M., Brown, D. G., & Neufeld, J. D. (2012).
Bolger, A. M., Lohse, M., & Usadel, B. (2014). Trimmomatic: A flexible trimmer for
PANDAseq: Paired-end assembler for illumina sequences. BMC Bioinformatics, 13(1),
Illumina sequence data. Bioinformatics, 30(15), 2114–2120.
31. https://doi.org/10.1186/1471-2105-13-31.
Chang, J.-H., Shim, Y. Y., Cha, S.-K., & Chee, K. M. (2010). Probiotic characteristics of
McLeod, A., Zagorec, M., Champomier-Vergès, M.-C., Naterstad, K., & Axelsson, L.
lactic acid bacteria isolated from kimchi. Journal of Applied Microbiology, 109(1),
(2010). Primary metabolism in Lactobacillus sakei food isolates by proteomic
220–230.
analysis. BMC Microbiology, 10(1), 120. https://doi.org/10.1186/1471-2180-10-
Chao, A. (1987). Estimating the population size for capture-recapture data with unequal
120.
catchability. Biometrics, 43(4), 783–791.
Moon, S. H., Kim, C. R., & Chang, H. C. (2018). Heterofermentative lactic acid bacteria as
Chazdon, R. L., Colwell, R. K., Denslow, J. S., & Guariguata, M. R. (1998). Statistical
a starter culture to control kimchi fermentation. LWT – Food Science and Technology,
methods for estimating species richness of woody regeneration in primary and
88, 181–188.
secondary rain forests of Northeastern Costa Rica. In Man and the Biosphere Series No.
Shannon, C. E. (1997). The mathematical theory of communication. 1963. In MD
Vol. 20. Forest biodiversity research, monitoring and modeling: conceptual background
Comput (pp. 306–317).
and old world case studies (pp. 285–309). Paris: Parthenon Publishing Group.
Song, H. S., Whon, T. W., Kim, J., Lee, S. H., Kim, J. Y., Kim, Y. B., … Roh, S. W. (2020).
Cheigh, H.-S., Park, K.-Y., & Lee, C. Y. (1994). Biochemical, microbiological, and
Microbial niches in raw ingredients determine microbial community assembly
nutritional aspects of kimchi (Korean fermented vegetable products). Critical Reviews
during kimchi fermentation. Food Chemistry, 318, 126481. https://doi.org/10.1016/
in Food Science and Nutrition, 34(2), 175–203.
j.foodchem.2020.126481.
Cho, J., Lee, D., Yang, C., Jeon, J., Kim, J., & Han, H. (2006). Microbial population
Steinkraus, K. H. (1997). Classification of fermented foods: Worldwide review of
dynamics of kimchi, a fermented cabbage product. FEMS Microbiology Letters, 257(2),
household fermentation techniques. Food Control, 8(5-6), 311–317.
262–267.
Toe, C. J., Foo, H. L., Loh, T. C., Mohamad, R., Abdul Rahim, R., & Idrus, Z. (2019).
Chun, B. H., Kim, K. H., Jeon, H. H., Lee, S. H., & Jeon, C. O. (2017). Pan-genomic and
Extracellular proteolytic activity and amino acid production by lactic acid bacteria
transcriptomic analyses of Leuconostoc mesenteroides provide insights into its
isolated from Malaysian foods. International Journal of Molecular Sciences, 20(7),
genomic and metabolic features and roles in kimchi fermentation. Scientific Reports,
1777.
7(1), 1–16.
Zheng, J., Wittouck, S., Salvetti, E., Franz, C., Harris, H. M. B., Mattarelli, P., & Lebeer, S.
Edgar, R. C. (2010). Search and clustering orders of magnitude faster than BLAST.
(2020). A taxonomic note on the genus Lactobacillus: Description of 23 novel genera,
Bioinformatics, 26(19), 2460–2461.
emended description of the genus Lactobacillus Beijerinck 1901, and union of
Edgar, R. C., Haas, B. J., Clemente, J. C., Quince, C., & Knight, R. (2011). UCHIME
Lactobacillaceae and Leuconostocaceae. International Journal of Systematic and
improves sensitivity and speed of chimera detection. Bioinformatics, 27(16),
Evolutionary Microbiology, 70(4), 2782–2858.
2194–2200.
Fadrosh, D. W., Ma, B., Gajer, P., Sengamalay, N., Ott, S., Brotman, R. M., & Ravel, J.
(2014). An improved dual-indexing approach for multiplexed 16S rRNA gene

9
 Update
Food Research International
Volume 156, Issue , June 2022, Page

DOI: https://doi.org/10.1016/j.foodres.2022.110946
 Food Research International 156 (2022) 110946

Contents lists available at ScienceDirect

Food Research International

journal homepage: www.elsevier.com/locate/foodres

Corrigendum

Corrigendum to “Effects of the main ingredients of the fermented food,

kimchi, on bacterial composition and metabolite profile” [Food Res. Int.
149 (2021) 110668]
Hye Seon Song a, 1, Se Hee Lee a, 1, Seung Woo Ahn a, Joon Yong Kim a, Jin-Kyu Rhee b, *,
Seong Woon Roh a, *
a
Microbiology and Functionality Research Group, World Institute of Kimchi, Gwangju 61755, Republic of Korea
b
Department of Food Science and Engineering, Ewha Womans University, Seoul 03760, Republic of Korea

The authors regret that the two research fund numbers (2018R1D1A1A09082921, 2018R1D1A1B07045349 and 2021R1F1A
(2021R1F1A1056188 and 321021031HD030) were omitted in the ac­ 1056188), and Ewha Womans University Research Grant of 2021, and
knowledgements section of the original article. High Value-added Food Technology Development Program, the Ministry
The corrected acknowledgements section is updated as follows: of Agriculture, Food and Rural Affairs (MAFRA) (321021031HD030),
Republic of Korea.
This research was supported by the World Institute of Kimchi funded
The authors would like to apologise for any inconvenience caused.
by the Ministry of Science and ICT (KE2101-2), Basic Science Research
Program of the National Research Foundation of Korea (NRF)

DOI of original article: https://doi.org/10.1016/j.foodres.2021.110668.

* Corresponding authors.
E-mail addresses: jkrhee@ewha.ac.kr (J.-K. Rhee), swroh@wikim.re.kr (S.W. Roh).
1
These authors contributed equally to this work.

https://doi.org/10.1016/j.foodres.2022.110946

Available online 28 February 2022

0963-9969/© 2021 The Author(s). Published by Elsevier Ltd. All rights reserved.