Identification of 2-hydroxyisocaproic acid production in lactic acid bacteria and evaluation of microbial dynamics during kimchi ripening

Lactic acid bacteria produce diverse functional metabolites in fermented foods. However, little is known regarding the metabolites and the fermentation process in kimchi. In this study, the culture broth from Leuconostoc lactis, a lactic acid bacterium isolated from kimchi, was analysed by liquid chromatography-tandem mass spectrometry and identified by the MS-DIAL program. The MassBank database was used to analyse the metabolites produced during fermentation. A mass spectrum corresponding to 2-hydroxyisocaproic acid (HICA) was validated based on a collision-induced dissociation (CID) fragmentation pattern with an identified m/z value of 131.07. HICA production by lactic acid bacteria was monitored and showed a positive correlation with hydroxyisocaproate dehydrogenases (HicDs), which play a key role in the production of HICA from leucine and ketoisocaproic acid. Interestingly, the HICA contents of kimchi varied with Leuconostoc and Lactobacillus content during the early stage of fermentation, and the addition of lactic acid bacteria enhanced the HICA content of kimchi. Our results suggest that HICA production in kimchi is dependent on the lactic acid bacterial composition.

resultant mass spectrum was analysed with the MassBank mass-spectral database using MS-DIAL software 22 . A mono-isotopic parent ion peak at m/z 131.07 and fragment masses at m/z 85.06 and 69.03 had the highest MS2 similarities and were identified as 2-hydroxyisocaproic acid (HICA). HICA is reported to be a leucine metabolite of Lactobacillus sp. We therefore characterized and analysed HICA production by lactic acid bacteria isolated from kimchi and monitored HICA production in relation to the bacterial composition during kimchi fermentation.

Results
Identification of 2-hydroxyisocaproic acid from lactic acid bacteria. To investigate the metabolites of lactic acid bacteria isolated from kimchi, we first attempted to identify the metabolites of Leuconostoc lactis using a UPLC system coupled to mass spectrometry (MS/MS). The untargeted acquisition of metabolome data was processed by the MS-DIAL program with the MassBank database 22 . In the resulting spectrum, one molecule had an m/z value of 131.07 with fragment ions of m/z 85.06 and 69.03; these had the highest MS2 similarities, exactly matching the known MS/MS fragmentation patterns for HICA in MassBank (http://www.massbank.jp/ en/ database.html) and METLIN (https://metlin.scripps.edu/metabo_advanced.php). To confirm the identity of the compound, collision-induced dissociation was performed to characterize the identified ion and compare the molecule with pure HICA (Sigma-Aldrich, St. Louis, MO, USA). The identified compound exhibited the same retention time and fragmentation pattern as pure HICA in product-ion MS mode (Fig. 1). The fragment ions at m/z 85.06 and 69.03 corresponded to neutral losses of 46.01 and 62.04 Da, respectively, from the precursor ion (m/z 131.07).

Analysis of hydroxyisocaproate dehydrogenases (HicDs).
To investigate HicDs, which catalyse the conversion to HICA, the amino acid sequences of HicDs were obtained from the protein sequence collection in UniProt (http://www.uniprot.org/). Leuconostoc lactis and Leuconoostoc mesenteroides each have 1 annotated HicD protein. Pediococcus pentosaceus and Lactobacillus sakei do not have any proteins corresponding to HicDs, while Lactobacillus brevis has two and Lactobacillus plantarum has 2-4 HicD proteins (Table 1).
Quantification of HICA and hicD gene expression. For quantification of HICA content, a bacterial culture extract was analysed by UPLC combined with electrospray ionization time-of-flight MS (UPLC-ESI-TOF-MS/MS). Quantification of HICA was achieved by comparing the peak area of l-norvaline (internal standard) with the areas of relevant peaks from extracted-ion chromatograms (internal standard, m/z 116 > 116; HICA, m/z 131 > 85). The HICA content of L. plantarum reached 526 ± 20.9 μg/ml at 48 h (P = 0.00078) after cultivation, whereas L. lactis and L. mesenteroides produced 153.1 ± 13.7 (P = 0.004) and 266.9 ± 5.9 μg/ml (P = 0.0002) at 12 and 48 h, respectively (Fig. 3a). HICA content began to increase at 6 or 12 h with the growth of lactic acid bacteria. L. plantarum possesses three HicD proteins, and it produced approximately two fold more HICA than L. lactis or L. mesenteroides (Fig. 3a). However, although L. brevis possesses two HicD proteins, its HICA production was the lowest among the four lactic acid bacteria studied. This may be because one of its HicD proteins (C2D5N4) exhibited low similarity to the other HicDs ( Fig. 2a and b). The HICA contents of L. plantarum and L. mesenteroides were positively correlated with the intracellular leucine content ( Fig. 3b and c). In addition, the HICA content of L. mesenteroides increased despite growth suppression at pH 5.5 (Fig. 3b).
hicD gene expression from lactic acid bacteria was measured using quantitative real-time PCR. Bacteria were cultivated in MRS broth in which the original pH of 6.2 was adjusted to 4.5-5.5 with lactic acid. hicD gene expression increased as bacterial growth increased or as pH decreased ( Fig. 3d and Supplementary Fig. S3). The three hicD homologues of L. plantarum all showed similar gene expression patterns. While L. brevis possesses two hicD genes, HMPREF0496_0286 (C2CYB4) did not exhibit an increase in gene expression with increasing bacterial growth ( Supplementary Fig. S3).
Next, we measured the HICA contents of kimchi obtained from a local market. The HICA contents of the kimchi samples began to increase after approximately 2 weeks and then decreased over a 3-4-week period. Kimchi HICA content ranged from 7.1 ± 0.1 to 22.6 ± 0.4 μg/ml during the period from 2 weeks to 4 weeks (Fig. 4a).
To investigate the association between HICA content and the kimchi microbiota, we examined the kimchi microbiota using metagenomic techniques and measured the acidity of each kimchi sample. As shown in Fig. 4b, acidity increased during kimchi fermentation, with all kimchi samples showing a similar pattern in which acidity increased from approximately 0.3% at the initial time point to approximately 1.2% after 4 weeks of fermentation. The overall HICA contents of kimchi increased (P < 0.05) after 1 week of fermentation, except for in one kimchi sample (no. 2), which nonetheless showed a similar acidity to those of the other kimchi samples. The microbiota of the kimchi was dominated by the Lactobacillus and Leuconostoc genera, in accordance with the presence of fermentation  Table 1. HICA production and HicD proteins of lactic acid bacteria. 1  To determine whether the addition of lactic acid bacteria contributes to the production of HICA in kimchi, we inoculated 10 7 CFU/g bacteria into kimchi. Kimchi inoculated with L. plantarum and L. mesenteroides showed higher contents of HICA (2.2-3.5 μg/ml) (P < 0.01) than control kimchi samples (1.1-1.5 μg/ml) (Fig. 4d). The pH and acidity of inoculated kimchi were not dependent on HICA content (Fig. 4e).

Discussion
In this study, we identified HICA from bacterial cultures of L. lactis, L. plantarum, L. brevis, and L. mesenteroides through an untargeted metabolomics approach. The amino acid derivative HICA (also known as leucic acid, 2-hydroxy-4-methylvaleric acid, and 2-hydroxy-4-methylpentanoic acid) is a protein-fermentation product of bacteria, such as lactobacilli. In addition, mammalian cells can metabolize HICA and use the metabolite for protein synthesis 25 . However, the generation of HICA in fermented vegetables and the lactic acid bacteria responsible for HICA production in fermented foods have not been investigated thoroughly.
A previous report showed that the Lactobacillus amylovorus DSM19280 (cereal isolate), Lactobacillus brevis R2Δ (porcine isolate), and Lactobacillus reuteri R29 strains produce HICA and characterized their antifungal activities 26 . HICA exhibits mild fungicidal or antibiotic activity against Candida, Aspergillus, Staphylococcus, and Fusobacterium 27, 28 and also shows activity against Enterococcus faecalis isolated from human teeth 29 . In a murine model, HICA attenuated inflammatory responses during Candida infection 25 . HICA can increase protein synthesis and improve muscle recovery after immobilization-induced atrophy 30 . The induction of protein synthesis is believed to occur through the activation of mammalian mTOR signalling, which is also activated during the innate inflammatory responses induced by bacteria, fungi, parasites, and viruses 30 .
HICA is formed by the transamination of leucine to 2-ketoisocaproic acid (KICA), followed by a reduction reaction of 2-KICA to 2-HICA, which is the end product of the leucine catabolism pathway. HicD is required for the latter reaction 27 . HICA is a typical constituent of human plasma, naturally circulating at a concentration of 0.25 ± 0.02 mmol/L 31 , and is found in muscle generally considered to have anti-catabolic activity 30 . HICA is also detectable in urine 32 and other biological fluids [33][34][35] . HICA can inhibit various matrix metalloproteinase enzymes that are responsible for degrading connective and protein tissues 36 .
HICA production during bacterial growth was measured in multiple-reaction monitoring (MRM) mode during LC-MS/MS analysis. HICA production ranged from 153.1 to 526 μg/ml in L. lactis, L. plantarum, L. brevis, and L. mesenteroides. HICA production was lower in L. sakei and P. pentosaceus 21.4 and 26.1 μg/ml, respectively), which do not have HicDs.  These results agree well with both the numbers and similarities of HicDs in the bacterial strains. HicD was not identified in P. pentosaceus and L. sakei, and these two lactic acid bacteria strains showed reduced HICA production during growth. However, HicD belongs to the family of lactate dehydrogenases (Table S1), suggesting basal expression of HICA even in the absence of HicDs.
Previously, changes in the microbiota during kimchi fermentation were well characterized, and it was shown that Lactobacillaceae and Leuconostocaceae are the dominant microbial families in kimchi, with Lactobacillus species present during the later stages of fermentation and Leuconostoc species found during the early stages of kimchi fermentation 6,37 . Even with varying degrees of acidity among the lactic acid bacteria, all cultures were acidic, and this increased in parallel with bacterial growth. Therefore, we studied HICA production in kimchi and microbial populations to evaluate HICA production during kimchi fermentation. The HICA content in kimchi began to increase after 1 week, with the exception of that in kimchi sample no. 2 (Fig. 4b). Kimchi samples no. 1 and 2 had higher HICA contents at the starting time point but had low acidity and were composed predominantly of Lactobacillus and Leuconostoc (Fig. 4c). Nonetheless, HICA contents were consistent with changes in the microbiota despite the low acidity. Moreover, HICA contents in kimchi with added lactic acid bacteria support this finding. HICA contents were highest in the kimchi samples with L. mesenteroides and L. plantarum (Fig. 4d).
Acidity is used as a standard to evaluate the progress of fermentation in kimchi and other fermented foods, as acidity increases with the duration of kimchi fermentation. Leucine and ketoisocarproic acid are the main substrates in the HICA catabolic pathway in bacteria. Isovaleric acid is the final product of leucine oxidization, while reduction promotes the final product of isocaproic acid 38 . Acidic stress in L. rhamnosus and L. reuteri results in an increase in ABC-type dipeptide/oligopeptide transporter (dpp/opp) expression, which plays an important role in amino acid transport 39,40 . A study into the L. casei amino acid uptake system suggested that leucine is transported by a proton drive motive force. Reductions in the proton motive force due to inhibitors or changes in pH decrease leucine transport 41 . In this study, we observed an increase in intracellular leucine, and this was positively correlated with HICA production (Fig. 3b,c). The fact that hicD expression was also increased at lower pH ranges supports the idea that the kimchi fermentation environment might be positively influenced by HICA production by lactic acid bacteria (Fig. 3d), especially by Lactobacillus and Leuconostoc during the initial kimchi fermentation process.
In conclusion, we found that the production of HICA as a leucine metabolite differed among lactic acid bacteria. These differences were mainly due to the core HicDs present in the lactic acid bacteria and to environmental stress during fermentation. These results also suggest the possible use of HICA as an indicator for fermented foods with lactic acid bacteria or related industrial processes.

Materials and Methods
Medium and bacterial culture conditions. All bacteria were isolated from homemade kimchi and identified by 16 S rDNA sequencing 42  Preparation of samples and standard solutions. Bacterial culture samples were collected at 0, 6, 12, 24, and 48 hours after incubation at 30 °C and were centrifuged at 10,000 × g for 10 min; l-norvaline was added as an internal standard (final concentration: 100 μg/ml). The samples were extracted using a Sep-Pak C18 Light cartridge (Waters, Milford, MA, USA). Finally, the collected samples were lyophilized in a Speed-Vac and resuspended with distilled water for LC-MS/MS analysis. To prepare standard solutions, a stock solution of HICA was diluted serially to concentrations of 100, 300, 500, 800, and 1000 μM in 10 ml MRS broth. l-Norvaline was used as an internal standard at a concentration of 100 μg/ml in 10 ml MRS broth or bacterial culture supernatant. HICA and the internal standard were obtained from Sigma Aldrich (St. Louis, MO, USA) and used to generate standard curves for subsequent quantification. All solvents were LC-MS grade and were purchased from J. T. Baker (Phillipsburg, NJ, USA).

UPLC-ESI-TOF-MS/MS conditions, HICA identification, and quantification. A TripleTOF 5600
plus instrument (SCIEX, Redwood City, CA, USA) coupled with an Acquity UPLC system (Waters) was used to characterize the metabolites and quantify the HICA contents of bacteria and kimchi. MS results were obtained at m/z 50-2000 in electrospray-negative mode with a spray voltage of −4.5 kV at a scan rate of 10 spectra/s. A reversed-phase column (Acquity UPLC BEH C18 column 2.1 × 100 mm, 1.7 μm particle size; Waters) was used to separate the compounds. The mobile phase consisted of distilled water (solvent A) and acetonitrile (solvent B) containing 10 mM ammonium acetate at a flow rate of 0.5 ml/min. The UPLC gradient program was as follows: (1) 95% solvent A from 0 to 0.5 min, (2) a linear gradient from 70% to 45% over 18.4 min, (3) 10% solvent A for 5 min. The total run time was 30 min per sample, and the injection volume was 2 μl. All experiments were performed in triplicate, and the data were processed using SCIEX PeakView 1.2 and MultiQuant 2.1 software (ABsciex). Peak picking and alignment were performed using MS-DIAL (ver. 1.98) 22 . Representative MS/MS spectra were exported in abf format for MS-DIAL, and compound identification was performed against MS/MS libraries including MassBank (MassBank_MSMS_Neg_Rev173_vs1) 44 and ReSpect (Respect_20120925_ESI_ Negative_MSMS) 45 . HICA from lactic acid bacteria and kimchi was quantified in MRM mode, using the following transitions: HICA, m/z 131.0 > 85.0; l-norvaline, m/z 116.
Comparison and analysis of metagenomic data. Metagenomic DNA was isolated from commercial kimchi and analysed by sequencing at Chunlab, Inc. (Seoul, Korea) using an Illumina MiSeq sequencing system (Illumina, San Diego, CA, USA) in accordance with the manufacturer's instructions.
The taxonomic classification for each read was determined using the EzTaxon-e database (http://eztaxon-e. ezbiocloud.net). The richness and diversity of samples were confirmed by Chao1 estimation and the Shannon diversity index at a 3% distance. To compare operational taxonomic units (OTUs) among samples, shared OTUs were identified by XOR analysis using the CL community program (Chunlab Inc., Seoul, Korea).