Food Chemistry 234 (2017) 416–424

Contents lists available at ScienceDirect

Food Chemistry

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

Process specific differential metabolomes for industrial gochujang types(pepper paste) manufactured using white rice, brown rice, and wheat

http://dx.doi.org/10.1016/j.foodchem.2017.04.1540308-8146/� 2017 Elsevier Ltd. All rights reserved.

⇑ Corresponding author at: Department of Bioscience and Biotechnology, KonkukUniversity, 120 Neungdong-ro, Gwangjin-gu, Seoul 05029, Republic of Korea.

E-mail address: chlee123@konkuk.ac.kr (C.H. Lee).1 These authors contributed equally to this work.

Yu Kyung Jang a,1, Gi Ru Shin a,1, Eun Sung Jung a, Sunmin Lee a, Sarah Lee b, Digar Singh a, Eun Seok Jang c,Dong Joo Shin c, Hye-jin Kim c, Hye Won Shin c, Byoung Seok Moon c, Choong Hwan Lee a,⇑aDepartment of Bioscience and Biotechnology, Konkuk University, Seoul 05029, Republic of KoreabNational Institute of Biological Resources, Environmental Research Complex, Incheon 404-708, Republic of Koreac Foods Research Institute, CJ CheilJedang Corp., Suwon 16495, Republic of Korea

a r t i c l e i n f o

Article history:Received 21 October 2016Received in revised form 29 March 2017Accepted 25 April 2017Available online 27 April 2017

Chemical compounds studied in this article:Valine (PubChem CID: 6287)LysoPC (18:2) (PubChem CID: 11005824)Serine (PubChem CID: 5951)myo-Inositol (PubChem CID: 892)Genistein (PubChem CID: 5280961)Daidzin (PubChem CID: 107971)Isoleucine (PubChem CID: 6306)Soyasaponin I (PubChem CID: 122097)Capsaicin (PubChem CID: 1548943)Dihydrocapsaicin (PubChem CID: 107982)

Keywords:GochujangMetabolomicsEnzymesPhysicochemical factorsProduction processes

a b s t r a c t

The metabolic perplexes for gochujang (GCJ) fermentative bioprocess, a traditional Korean pepper paste,has largely remain equivocal for preparative conditions and raw material (RM) additives exacerbating itscommercial standardization. Herein, we outlined a differential non-targeted metabolite profiling forthree GCJ (white rice-WR; brown rice-BR; wheat-WT) under varying processing steps (P1 – fermentation;P2 – meju addition; P3 – ripening; and P4 – red pepper addition). We correlated the process specificmetabolomes with corresponding physicochemical factors, enzymatic phenotypes, and bioactivities forGCJ-types. The P1 was characterized by a uniform increase in the levels of RM-derived lysoPCs. In con-trast, P2 was observed with proportionally higher levels of meju-released isoflavones and soyasaponinsin WR-GCJ, followed by BR and WT-GCJ. The P3 involved a cumulative increase in primary metabolitesin all GCJ samples except lower organic acid contents in WT-GCJ. The pepper derived flavonoids and alka-loids were selectively increased while P4 in all GCJ-types.

� 2017 Elsevier Ltd. All rights reserved.

1. Introduction

Fermented foods have traditionally been the part and parcel ofethnic dietary culture ubiquitously in human societies with theirown peculiar characteristics, owing to the artisanal productionmethods evolved in different geo-sociological regions (Cvetkovicet al., 2015). In recent years, an overwhelming number of studieshave explicitly highlighted the various functional perspectives offermented foods despite their nutritional aspects viz., anti-obesity, antitumor, and anticancer (Cho et al., 2013). Gochujang, a

toothsome piquant red pepper soybean paste, is a traditional Kor-ean fermented condiment blended with glutinous grains (rice orwheat), pulverized fermented soybeans (meju), and pepper powder(Shin, Kim, Choi, Lim, & Lim, 1996). In recent years, a number ofreports have particularly highlighted the health effects of gochu-jang viz., anti-obesity, anti-cancer, or anti-metastatic, owing to anarray of functional metabolites derived from capsaicin, soy isofla-vones, and soyasaponins (Kim et al., 2005; Park, Kong, Jung, &Rhee, 2001; Shin et al., 2016). Buoyed by the globalization of theKorean fermented foods, gochujang have gained the consumersattention worldwide. The fact is evident by the recorded steep con-sumption and production statistics for commercial gochujang withan overall annual trade and exports of worth over $24 million and$28 million, respectively (Korea Agro-Fisheries & Food Trade

Fig. 1. A schematic sketch for the various processes involved in the industrialproduction of gochujang (GCJ) with different raw materials, RM (WR-white rice; BR-brown rice; WT-wheat). The various processes involved are indicated as, Process 1(P1): fermentation; Process 2 (P2): addition of meju powder (meju Addition);Process 3 (P3): ripening of semi-finished products (Ripening); Process 4 (P4):addition of red pepper powder and etc. (red pepper addition).

Y.K. Jang et al. / Food Chemistry 234 (2017) 416–424 417

Corporation, 2015). Owing to its commercial trends, the qualitycontrol parameters for characteristic taste, aroma, and flavor dur-ing the manufacturing processes of gochujang are also being accen-tuated (Hong et al., 2012). Traditionally, gochujang has been madeusing artisanal fermentation practices withmeju powder as naturalinocula. However, the artisanal gochujang manufacturing requiresa long fermentation time (�3 years) and seldom carries an unto-ward risk of microbial food intoxication (Nout, 1994). Moreover,the perplexing microbial community dynamics further turns itsreal-time production monitoring into an arduous task with poorlycontrolled aesthetics for end products (Baek et al., 2010). In con-trast, the optimized production of commercial gochujang couldhelp in formulating a more standardized product under reducedproduction time and controlled parameters (Yu et al., 2012). Inaddition, the quality of gochujang is significantly influenced bythe type and ratio of ingredients as well as the ripening conditions(Kim, Han, & Kim, 2010). Therefore, a detailed insight of the gochu-jang manufacturing through metabolic delineation of its taste,aroma, nutritional, and functional aspects could potentially pavethe way for standardization of its optimal fermentative production.

Metabolomics, which encompasses a holistic analysis ofchanges in low molecular weight (<1800 Da) metabolites in a cell,tissue, or an organism under varying physiological conditions, isprimarily carried out using hyphenated mass spectrometry orspectroscopic methods (Zhou, Xiao, Tuli, & Ressom, 2012).Quintessentially, the food metabolomics involves the analyses ofnutritional contents, quality biomarkers, and health effects of foodssubjected to a bioprocess (Oh, Jang, Woo, Kim, & Lee, 2016). Inrecent times, a variety of traditional fermented foods viz., cheong-gukjang (Oh et al., 2016), doenjang (Lee, Lee, et al. (2014)), and thesoy fermented blends of Cudrania tricuspidata or Lonicera caerulea(Suh et al., 2016) etc. have been studied for their metabolomicimplications in conjunction with quality parameters (Kim et al.,2010).

The present investigations involves the comparative mass spec-trometry (MS) based metabolite profiling towards delineating theprocess specific metabolic alterations in a four step commercialgochujang production with varying raw materials (RM). We furthermade correlations among the process specific significantly discrim-inant metabolites and the product phenotypes viz., antioxidantactivity along with total phenolic and flavonoid contents (TFC).Additionally, the physicochemical factors cum quality parametersviz., pH, titratable acidity (TA), amino-type nitrogen contents, amy-lase activity, and a-glucosidase activity were also evaluated toinfer the effects of varying grain blends in different processingsteps.

2. Materials and methods

2.1. Chemicals and reagents

All the chemicals and reagents used in this study were of ana-lytical grade. Acetonitrile, water, and methanol were purchasedfrom Fisher Scientific (Pittsburgh, PA, USA). Potassium dihydrogenphosphate, dipotassium hydrogen phosphate, sodium carbonate,sodium hydroxide, sodium chloride, and diethylene glycol werepurchased from Junsei Chemical Co., Ltd. (Tokyo, Japan). Trichlor-oacetic acid was purchased from Merck Millipore Co. (Darmstadt,Germany). Methoxyamine hydrochloride, N-methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA), pyridine, potassiumpersulfate, 2,20-azinobis (3-ethylbenzothiazoline-6-sulfonic acid)diammonium salt (ABTS), Folin–Ciocalteu’s phenol reagent, solublestarch, potassium sodium tartrate tetrahydrate, acetic acid,sodium acetate, 4-nitrophenyl-a-D-gluco-pyranoside, 1 N sodiumhydroxide solution, formaldehyde solution, formic acid,3,5-dinitrosalicylic acid (DNS), and the standard compounds

6-hydroxy-2,5,7,8-tetramethylchromane-2-carboxylic acid (Trolox),naringin, gallic acid, q-nitrophenol, norvaline, formononetin, wereobtained from Sigma Chemical Co. (St. Louis, MO, USA).

2.2. Sample preparations

The three different commercial gochujang (GCJ) types based ondifferent grain blends i.e., white rice (WR), brown rice (BR), andwheat (WT) were procured from ’CJ Cheiljedang Corporation’Seoul, South Korea. Each of the GCJ types were collected from fourmain processes of its industrial production. As shown in Fig. 1, pro-cess 1 (P1 – fermentation) involved the fermentation of steamedraw material (RM) viz., WR, BR, or WT using Aspergillus oryzae at35 �C for 36 h and finally resulted in koji (kJ) formation i.e., 15–25% of final products. The process 2 (P2 – meju addition) was car-ried out by addingmeju powder (crushed soybeans fermented withA. oryzae, 1–5% of final products) to kJ to obtain a semi-finishedproduct (SP1) before ripening. Following, the process 3 (P3 – ripen-ing) involved the ripening of SP1 to another semi-finished product(SP2) through incubation at 30 �C for 21 days. Finally, the process 4(P4 – red pepper addition, 10–20% of final product) was accom-plished by adding the optimal proportions of red pepper powder,garlic, salt, sugar syrup, and water to obtain the desired endproduct for each of the commercial GCJ-types (WR-GCJ, BR-GCJ,WT-GCJ).

2.3. Sample extractions and preparation for analyses

The WR-GCJ, and BR-GCJ samples from each process werefreeze-dried for 3 days, pulverized, and extracted with 80% metha-nol (100 mg/mL) at 25 �C through vigorous shaking for 24 h using aTwist Shaker (Biofree, Seoul, Korea). The equivalent amount of pre-treated WT-GCJ samples in 80% methanol were extracted throughvigorous shaking (30 Hz/s) for 10 min using a mixer mill (RetschMM 400, GmbH & Co, Germany). Sample extracts was performedin three biological replicates. All the sample extracts were cen-trifuged (5000 rpm, 4 �C, 10 min, Hettich Zentrifugen, Universal320R, Germany), and the decanted supernatants were filteredthrough a 0.22 lm polytetrafluoroethylene (PTFE) filter. The

418 Y.K. Jang et al. / Food Chemistry 234 (2017) 416–424

filtered supernatants were concentrated using a speed vacuumconcentrator (Modulspin 31, Biotron, Korea) followed by dissolu-tion in 80% methanol. Norvaline (10 L, 0.1 mg/mL), used as aninternal standard, was added to 190 lL of each dissolved extractsample, which were then dried again in a speed vacuum concentra-tor. The samples preparation for gas chromatography-time offlight-mass spectrometry (GC-TOF-MS) analysis involved a two-step derivatization reaction. First, the sample methoxymationwas carried out by dissolving dried extracts in methoxyamine–HCl (100 lL, 20 mg/mL in pyridine) and incubation at 30 �C for90 min. Secondly, the methoxymated samples were silylated byadding MSTFA (100 lL) and incubating the reaction mixture at37 �C for 30 min. The sample extracts (90 lL) for ultrahigh-performance liquid chromatography-linear trap quadrupole-electrospray ionization-ion trap mass spectrometry/mass spec-trometry (UHPLC–LTQ-ESI-MS/MS) analysis, were added with for-mononetin (10 lL, 0.4 mg/mL) as an internal standard. Analysesof the sample extracts were performed in triplicate.

2.4. GC-TOF-MS analysis for WR–GCJ, BR–GCJ, and WT-GCJ sampleextracts

An Agilent 7890 gas chromatography system (AgilentTechnologies, Palo Alto, CA, USA), equipped with an Agilent 7693

Amino type nitrogen ð%Þ ¼ ðVolume of 0:1 N NaOH ðmLÞ � Volume of retitration 0:1 N NaOH ðmLÞÞ � 1:4 ðmg=mlÞ � 5df

Homogenized samples ðmgÞ� �

� 100

autosampler and Pegasus� High-Throughput-TOF-MS (LECO, St.Joseph, MI, USA), was used for GC-TOF-MS analysis of GCJ sampleextracts. The GC-TOF-MS analysis and operational parameters usedin this study were adapted from Oh et al., 2016.

2.5. UHPLC-LTQ-ESI-IT-MS/MS analysis for WR-GCJ and BR-GCJsample extracts

A Thermo Fischer Scientific LTQ ion trap mass spectrometerwas employed with an electrospray interface (Thermo FischerScientific, San José, CA), DIONEX UltiMate 3000 RS Pump, RSAutosampler, RS Column Compartment, and RS Diode ArrayDetector (Dionex Corporation, Sunnyvale, USA). The conditionsand operational parameters for UHPLC-LTQ-ESI-IT-MS/MS analysiswere adjusted according to the method described by Oh et al.,2016.

2.6. UPLC-Q-TOF-MS analysis of WT-GCJ sample extracts fromdifferent processes

UPLC-Q-TOF-MS analysis was performed using a UPLC system(Waters Corp., Milford, MA, USA) combined with a Q-TOF PremierMS (Waters Micromass Technologies, Manchester, UK) system,coupled with a UV detector and an autosampler. For GCJ sampleextracts representing each process in its industrial production,5 L was injected into an ACQUITY BEH C18 column(100 mm � 2.1 mm i.d., 1.7 lm,Waters Corp.). The gradient mobilephase consisted of water (solvent A) and acetonitrile (solvent B),both containing 0.1% formic acid. The mobile phase gradient wasinitially maintained at 5% B for 1 min, then raised to 100% B over10 min at a flow rate of 0.3 mL/min, held at 100% B for 2 min,and then finally decreased to 5% B over 1 min. The total analysis

run time was 14 min, including column re-equilibration to the ini-tial conditions. The metabolites were analyzed in both positive andnegative ion modes under the following operating conditions: conevoltage, 40 V; capillary voltage, 2.5 kV; mass range, m/z 100–1000;and source and desolvation temperatures of 100 and 300 �C,respectively. Capillary voltages were set at �2.3 kV and +2.5 kVin negative and positive ionization modes, respectively. All peakdata were analyzed with Mass Lynx software (Version 4.1, WatersCorp.).

2.7. Determination of physiochemical factors – pH, titratable acidity,and amino-type nitrogen contents

The GCJ (5 g) sample replicates (n = 3) for each process in itsindustrial production were homogenized with distilled water(25 mL) and their pH was measured using a pH meter (Thermo,USA). Further, the titratable acidity was estimated by titratingthe homogenized samples using 0.1 N NaOH solution to a set pHlevel of 8.4. The amino-type nitrogen contents were measure byadding 36% formaldehyde solution (20 mL) to the titratedsamples, followed by sample re-titration to pH 8.4 using 0.1 NNaOH solution. The quantitative estimation for amino-type nitro-gen contents in the samples was performed using the followingformula;

All the results of quantitative analyses for three GCJ-types (WR-GCJ, BR-GCJ, WT-GCJ) representing each process (P1, P2, P3, P4)while industrial production were expressed as an average valuesupplemented with standard deviations.

2.8. Determination of amylase and a-glucosidase activities

The amylase and a-glucosidase activities were estimated usingthe modified method adapted from Ghosh et al. (2015). The crudeenzymes from process specific GCJ samples (10 g) were extractedwith water (90 mL) under vigorous agitation using a shaking incu-bator (30 �C, 120 rpm, 1 h). Subsequently, the samples were cen-trifuged (5000 rpm, 4 �C, 10 min), and the supernatants werefiltered through a 0.2 lm polytetrafluoroethylene (PTFE) filter fol-lowed by enzymatic activity analysis. The data for enzymatic activ-ities determined for three GCJ-types (WR-GCJ, BR-GCJ, WT-GCJ)representing each process (P1, P2, P3, P4) while industrial produc-tion were expressed as an average value supplemented with stan-dard deviations.

2.9. Determination of total polyphenol and total flavonoid contents

Total polyphenol contents (TPC) and the total flavonoids con-tents (TFC) assays were performed according to the Suh et al.,2016 method. In TPC assay, each sample extract (20 mL) wasreacted with 0.2 N Folin–Ciocalteu reagent (100 lL) for 5 min inthe dark. Then, 7.5% NaCO3 (80 lL) was added, and absorbancewas measured at 750 nm using a spectrophotometer. Standardcurves were made in the linear range of 7.81 ppm and 500 ppmfor gallic acid equivalents. The results were expressed in ppm ofgallic acid per milligram of each sample extract.

Y.K. Jang et al. / Food Chemistry 234 (2017) 416–424 419

The TFC assay, each sample extract (20 lL) was mixed with 1 NNaOH (20 lL) and 90% diethylene glycol (180 lL). The reactionmixture was incubated in dark for 60 min at room temperatureand the absorbance was recorded at 405 nm using a spectropho-tometer. The standard curve was linear between the concentra-tion range 6.25 ppm and 200 ppm of naringin equivalents, whilethe results were expressed in ppm of naringin per milligram ofsample extracts. All experiments were performed in triplicate,using the same sample extracts as those in mass spectrometryanalysis.

The TPC and TFC data for three GCJ types in each process wereexpressed as an average value with standard deviations.

2.10. Determination of antioxidant activity

The ABTS assay was performed according to the methoddescribed by Suh et al., 2016, with some modifications. The7 mM ABTS stock solution was dissolved in 2.45 mM potassiumpersulfate solution, and the mixture was stored for 24 h in darkat 4 �C. The solution was diluted until the absorbance reached0.7 ± 0.02 at 750 nm, measured using a spectrophotometer(SpectronicGenesys 6; Thermo Electron, Madison, WI, USA). Eachsample extract (20 lL) was reacted with diluted ABTS solution(180 lL) for 7 min at room temperature in dark, followed byabsorbance measurement at 750 nm. A standard curve wasmade in the linear range, 0.0625 mM–2 mM, Trolox equivalents,with results were expressed in lmol of Trolox equivalents (TE)per milligram of each sample. All assays were performed intriplicate, using the same sample extracts as those in mass spec-trometry analysis. The ABTS assay data for three GCJ-types in eachprocess were expressed as an average value with standarddeviations.

2.11. Data processing and multivariate statistical analysis

GC-TOF-MS raw data files were converted to computabledocument form (⁄.cdf) using LECO chroma TOFTM software, andUPLC-Q-TOF-MS data were converted into a NetCDF files (.cdf)using MassLynx DataBridge (Version 4.1, Waters Corp.). Addition-ally, the accurate masses and elemental compositions were calcu-lated using the MassLynx software. UHPLC-LTQ-ESI-IT-MS/MS rawdata files were converted using Thermo Xcalibur software(Version 2.1; Thermo Fisher Scientific Inc., USA). Acquired .cdfformatted data were subjected to preprocessing alignment usingthe MetAlign software package (http://www.metalign.nl). Afteralignment, the resulting peak list was obtained as a .txt file, whichwas exported into Microsoft Excel (Microsoft, Redmond, WA,USA). The Excel file included the corrected peak retention times(min), peak areas, and corresponding mass (m/z) data for furtheranalysis. The pre-processed data were subjected to multivariatestatistical analysis using SIMCA-P+ 12.0 software (Version 12.0,Umerics, Umea, Sweden) and the trends of metabolite variations(primary and secondary metabolites) in GCJ-types from industrialprocessing steps were determined using principal componentanalysis (PCA) and partial least-square discriminant analysis(PLS-DA). Selected metabolites with variable importance in theprojection (VIP) values of >0.7 and a p-value < 0.05 were consid-ered significantly discriminant among the samples. For TPC, ABTS,and TFC, differences were tested by analysis of variance (ANOVA)and Duncan’s multiple range tests, using PASW Statistica 18 (SPSSInc., Chicago, IL, USA). A heat map was produced using MeVsoftware version 4.8 [multiple array viewer, TM4: MicroarraySoftware Suite. Available online: http://www.tm4.org/ (accessedon September 25, 2014)].

3. Results

3.1. Effects on physiochemical factors and secreted enzymes in GCJ-types during industrial production processes

The average values for the physiochemical factors i.e., pH levels,titratable acidity, and amino-type nitrogen as well as the amylaseand a-glucosidase production were evaluated to estimate the qual-ity characteristics for GCJ-types according to the processes (P1, P2,P3, and P4) while industrial production (Fig. 2). The pH levels fellsharply, approximately pH 6.5 to pH 4.8, during P1 (Fermentation)and then maintained as such in the range of pH 4.5–4.8 in subse-quent processes (P2–P4) (Fig. 2A). However, similar trends wereobserved for variations in titratable acidity and amino-type nitro-gen contents with an observed increase until P3 and a fall thenafter (Fig. 2B and C). Especially, the values of amino-type nitrogencontents were observed significantly higher during ripening (P3)process. On the other hand, the amylase and a-glucosidase activi-ties were observed maximum during P2 and P1, respectively, fol-lowed by a gradual decreased (Fig. 2D and E).

3.2. Metabolite profiling of GCJ-types during industrial productionprocesses

The metabolite profiling through PCA analysis exhibited similarpatterns for metabolite variations based on GC-TOF-MS datasetsfor GCJ extracts i.e., WR-GCJ, BR-GCJ, and WT-GCJ from differentprocesses (P1, P2, P3, P4) in its industrial production (Fig. 3A,B, and C). Intriguingly, PCA analyses based on UHPLC-LTQ-ESI-IT-MS/MS data for WR-GCJ and BR-GCJ (Fig. 3D and E) as well asUPLC-Q-TOF-MS data for WT-GCJ (Fig. 3F), also exhibited a similartendency for metabolites irrespective of the processes. The PCAscore plots for WR-GCJ, BR-GCJ, and WT-GCJ metabolomic datasetsindicated clear grouping patterns for each state of production i.e.,raw materials (RM), koji (kJ), semi-finished products before ripen-ing (SP1), semi-finished products (SP2), and final GCJ. Except forthe PCA analysis towards UPLC-Q-TOF-MS datasets for WT-GCJ(Fig. 3F), all other PCA plots showed a distinct segregation for firstprocess (P1) characterized by the transition of RM to kJ while fer-mentation. In contrast, metabolite variations for rest of the pro-cesses viz., P2 (kJ? SP1), P3 (ripening), and P4 (SP2? GCJ)altered along PC2 for the process additives including meju andred pepper powder, in kJ and SP2, respectively. The significantlydiscriminant metabolites in three GCJ-types according to the pro-duction process were determined using the VIP values (VIP > 0.7)and p-values (p < 0.05) along PLS1 and PLS2 for correspondingPLS-DA model. The overall identified discriminant metabolites foreach GCJ types were divided into nine categories viz., amino acids,organic acids, fatty acids, sugars & sugar alcohols, lysoPCs, isofla-vones, soyasaponins, flavonoids, and alkaloids. These metaboliteswere identified by comparing retention times, MS/MS fragmentpatterns, molecular weights, standard compounds, an in-houselibrary (Lee, Kim, Liu, Oh, & Lee, 2005), NIST (National Institute ofStandards and Technology mass search version 2.0, 2011, USA)library, and the published literatures (Table S1–S6).

3.3. Metabolite correlations and associated bioactivities for three GCJ-types based on different processes while industrial production

The pair wise correlations among the significantly discriminantprimary and secondary metabolite for GCJ-types (WR-GCJ, BR-GCJ,WT-GCJ) were visualized using the heat maps (Fig. 4) during theprocesses of industrial production. Further, the specific alterationsin metabolites according the processes in industrial GCJ production

Fig. 2. The levels of physicochemical factors; (A) pH levels, (B) titratable acidity, (C) amino-type nitrogen, along with secreted enzymes; (D) amylase activity, (E) a-glucosidase activity for GCJ-types for different production processes. All the data were presented as the mean value (n = 3) with standard deviations (±SD) for GCJ-types (WR-GCJ, BR-GCJ, WT-GCJ) corresponding to different processes (P1, P2, P3, P4) in industrial productions. The various processes involved are indicated as, Process 1 (P1):fermentation; Process 2 (P2): addition of meju powder (meju Addition); Process 3 (P3): ripening of semi-finished products (Ripening); Process 4 (P4): addition of red pepperpowder and etc. (red pepper addition).

Fig. 3. The PCA score plots derived from GC–TOF–MS (A, B, and C), UHPLC–LTQ–ESI–IT–MS/MS (D and E), and UPLC-Q-TOF-MS (F) data sets for different processes symbolizedas: P1 (fermentation: +), P2 (meju addition: s), P3 (ripening: ▲), and P4 (red pepper addition: j). The plots representing different processes for GCJ types are indicated withdifferent colors; WR-GCJ , BR-GCJ , and WT-GCJ .

420 Y.K. Jang et al. / Food Chemistry 234 (2017) 416–424

are listed in Table 1. Additionally, the discriminative metabolitesidentified through various hyphenated MS-based analytical meth-ods are shown in Supplementary Tables S1–S6. In process 1, RMturned koji through A. oryzae mediated fermentation followed by

the generation of primary metabolites viz., amino acids (valine, iso-leucine, serine, threonine, etc.), organic acids (acetic acid, etc.),fatty acids (propanoic acid, etc.), and sugars & sugar alcohols(myo-inositol, sucrose, etc.). In case of WR-GCJ and BR-GCJ, espe-

Fig. 4. Heat map representations for changes in the relative levels of significantly discriminant primary and secondary metabolites among (A) WR-GCJ, (B) BR-GCJ, and (C)WT-GCJ, according to the processes while industrial GCJ production. The results of in vitro assays for (D) antioxidant activity test (ABTS), (E) total polyphenol contents (TPC),and (F) total flavonoid contents (TFC) were presented as the mean values (±SD) for GCJ-types (WR-GCJ, BR-GCJ, WT-GCJ) corresponding to different processes (P1, P2, P3, P4)in industrial productions.

Table 1The alteration of primary and secondary metabolites in gochujang (fermented red pepper paste) according to production process.

Process Increase Decrease

P1 (Fermentation) � Amino acids (Valine, Isoleucine, Serine, Threonine, etc)� Organic acids (Acetic acid, etc.)� Fatty acid (Propanoic acid, etc.)� Sugars & sugar alcohols (myo-Inositol, Sucrose, etc.)� Lyso PCs

–

P2 (meju Addition) � Isoflavones (Daidzin, Genistin, Daidzein, etc)� Soyasaponins (I, II, III, IV, V)

� Amino acids (Isoleucine, etc.)� Organic acid (Citric acid, etc.)� Fatty acid (Propanoic acid)� Sugars & sugar alcohols (Xylitol, Sucrose, etc.)� Lyso PCs

P3 (Ripening) � Amino acids (Alanine, Isoleucine, Glycine, Aspartic acid, Serine,pyroglutaric acid, Valine, Proline, Threonine, GABA, Glutamic acid, etc.)

� Organic acid (Lactic acid, Acetic acid, Malic acid, Citric acid,Succinic acid, etc.)

� Sugars & sugar alcohols (Xylose, Glucose, Glucopyranoside, myo-Inositol, etc.)� IsoFlavones (Genistin, Daidzin, Daidzein, Glycitein, Genistein)� Soyasaponins (I, II, III, IV)

� Lyso PCs

P4 (red pepper Addition) � Organic acid (Citric acid, etc.)� Sugars & sugar alcohols (Maltose, etc.)� Flavonoids (Luteolin 7-o-apiosyl-glucoside, Apiin, Apigenin-glucoside, etc.)� Alkaloids (Capsaicin, dihydrocapsaicin)

� Amino acids (Alanine, Valine, Threonine, Aspartic acid, etc.)� Fatty acid (Propanoic acid, etc.)� Sugars & sugar alcohols (myo-Inositol, etc.)� Soyasaponins (I, II, III, IV, V)

The criteria for selection is shown as in common metabolite which increased or decreased in two or more red pepper paste among WR, BR, and WT.

Y.K. Jang et al. / Food Chemistry 234 (2017) 416–424 421

422 Y.K. Jang et al. / Food Chemistry 234 (2017) 416–424

cially, the levels of lysoPCs were significantly increased while P1(late RM fermentation) and P2 (early meju addition)(Fig. 4A and B). In contrast, the lysoPCs levels were not selectivelyspiked for WT-GCJ while the fermentation phases (Fig. 4C). Theprocess P2 (meju addition) was followed by the specific rise inthe levels of soybean-originated metabolites viz., isoflavones (daid-zin, genistin, daidzein, glycitein, genistein) for all GCJ-types, andsoyasaponins (V, I, II, III, IV) except WT-GCJ. The largest numbersof metabolites were altered during P3 (ripening process) with asoybean-originated metabolites increased briskly. Additionally,the primary metabolites levels viz., amino acids (alanine, isoleu-cine, glycine, aspartic acid, serine, pyroglutamic acid, valine, pro-line, threonine, GABA, glutamic acid, etc.), organic acids (lacticacid, acetic acid, malic acid, citric acid, succinic acid, etc.), and sug-ars & sugar alcohols (xylose, glucose, glucopyranose, and myo-inositol, etc.) increased varyingly during P3 for all GCJ-types. How-ever, lysoPCs levels were significantly fell in all cases. In P4, thefinal step in GCJ production (red pepper addition), the contentsof red pepper-originated flavonoids and alkaloids, including lute-olin 7-O-apiosyl glucoside, apiin, apigenin glucoside, capsaicin,and dihydrocapsaicin were selectively increased, while those forisoflavones and soyasaponins were decreased. The relative levelsof metabolite classes viz., amino acid, organic acid, fatty acid, sugar& sugar alcohols, lysoPC, soyasaponin, isoflavone, flavonoid, and

Fig. 5. Box and whisker plots for nine metabolite classes outlining the differential metaaveraged for GCJ-types (WR-GCJ, BR-GCJ, and WT-GCJ) according to the processes and eameju powder (meju addition); process 3 (P3): ripening of semi-finished products (Ripen

alkaloid were further represented using the box and whisker plots(Fig. 5). The trends for metabolite classes in GCJ-types from differ-ent processes were found in agreement to those observed in heatmap analysis.

The mean alterations in bioactivity phenotypes while differentprocesses of industrial manufacturing for three GCJ-types wereevaluated through ABTS assay for antioxidant activity (Fig. 4D),TPC assay for total polyphenol contents (Fig. 4E), and TFC assayfor total flavonoid contents (Fig. 4F). In congruence to the metabo-lite profiles, the observed bioactivities were also varied during thedifferent processes of GCJ production. Specifically, the ABTS assaydetermined antioxidant activities were observed highest whilethe late P2 (meju addition) and early P3 (ripening) followed by aslight decrease in later stages. However, a marginal increase inantioxidant activities were again observed after adding red pepperpowder in P4. The TPC activity was specifically spiked while P3(ripening step), whereas the TFC activity was gradually increasedfrom P1 to P4.

4. Discussion

In the present study, we adopted a metabolomic approach todelineate the quality characteristics for three GCJ-types (WR-GCJ,BR-GCJ, and WT-GCJ) according to the processing steps viz., P1

bolic alterations according to the processes in GCJ manufacturing. All values werech process was indicated; process 1 (P1): Fermentation; process 2 (P2): addition ofing); process 4 (P4): addition of red pepper powder and etc. (red pepper addition).

Y.K. Jang et al. / Food Chemistry 234 (2017) 416–424 423

(fermentation), P2 (meju addition), P3 (ripening), and P4 (red pep-per addition) while industrial production. A variety of metabolitesand nutrients were dynamically altered during the different pro-duction processes owing to the different raw materials and blendsof additive materials subjected to microbial fermentations. Thealtered metabolomes in each GCJ-types were also manifested intheir bioactivity phenotypes (TPC, ABTS, TFC) as well as the qualityparameters (pH, titratable acidity, amino-type nitrogen contents,amylase activity, and a-glucosidase activity).

The P1 is primarily characterized by the A. oryzae mediated fer-mentative conversion of RM to kJ (koji). In this process, a plethoraof discriminative metabolites were observed depending upon theRM contents viz., WR, BR, or WT, which were significantly alteredduring the later processes. Table 1 summarizes the significantlyaltered metabolites with their relative levels increased ordecreased during the different processes of GCJ production. In P1,the relative levels of various classes of discriminant metabolitesviz., amino acids, organic acids, fatty acids, sugars & sugar alcohols,and lysoPCs were variously increased among the different GCJ-types (Fig. 4). The observed inception of primary and secondarymetabolite synthesis during P1 was consistent with a steep fallin pH levels coupled with a gradual rise in titratable acidity andamino-type nitrogen contents for the GCJ sample extracts(Fig. 2A, B, and C). The successive increase in an average antioxi-dant activity during P1 (Fig. 4D) can be directly linked with an ele-vation in the levels of lysoPCs (Fig. 5) as well as few amino acidsviz., valine, isoleucine, and GABA (Itoh, Kawashima, & Chibata,1975, and Chang, Han, Wight, & Chait, 2006). In case of WR-GCJand WT-GCJ, the overall levels of discriminant primary metaboliteswere significantly altered for the increased activity of hydrolyticenzymes secreted by Aspergillus species (Lee et al., 2016). In con-trary, the levels of primary metabolites while P1 in BR-GCJ werenot significantly altered owing the probable morphological cessa-tion of hydrolytic activity by rice bran in BR substrate(Moongngarm & Saetung, 2010). As indicated in correspondingheat maps (Fig. 4) and box plots (Fig. 5), the meju powder additionto koji during P2 was followed by a significant increase in the levelsof soybean-originated metabolites, such as isoflavones and soyas-aponins viz., daidzin, genistin, daidzein, and soyasaponins (I, II,III, IV, V). Functionally, soyasaponin compounds are known to havea number of functional effects viz., neuroprotective, hypocholes-terolemic, and anti-hepatotoxic (Hong et al., 2013; Lin, Meijer,Vermeer, & Trautwein, 2004; Yang, Dong, & Ren, 2011). Concur-rently, TFC and antioxidant activity were also observed higherowing to the significant levels of soy derived metabolites in lateP2 stages, as reported earlier by Lee, Lee, et al. (2014). Intriguingly,the addition of meju powder, except for WT-GCJ, was followed by asignificant decrease in the levels of primary metabolites andlysoPCs. On the other hand, the onset of P2 also witnessed a steepdrop in the pH levels (�6.5–4.5) owing to the release of organicacids during late P1 (fermentation). Further, the pH levels wereobserved varying in a narrow range (�4.4–5.0) in consecutive pro-cesses i.e., P2 to P4 on account of the higher levels of free aminoacids (Fig. 2A) (Kim & Yi, 2010). In food processing, the pH levelsare of vital importance for enhanced shelf-life of end products(Salmond, Kroll, & Booth, 1984 and Kim & Yi, 2010). Salmond andcoauthors have suggested that the growth of harmful E. coli stainsin food is effectively reduced below the pH 6.4, providing a mildacidic inhibitory environment. Therefore, the maintained pH valueis one of the important quality parameter for industrial productionof fermented products. Notwithstanding, the inoculation of A. ory-zae conidia and fermented soy supplement (15–25% of end productGCJ) while P1 and P2, respectively, we didn’t observed the Aspergil-lus specific metabolites. We presume that the fungal originatedmetabolites were not selected as significantly discriminant(VIP > 0.7, p < 0.05) owing to their marginal and uniform produc-

tion in all GCJ-types. Moreover, the meju (15–25% of end products)addition in P2 also supplemented the bioprocess with maximumproportions of soy derived metabolites. Hence, one can assumethe higher and discriminant abundance of mostly the substrate(soy or cereals) derived metabolites released through its briskenzymatic hydrolysis in the subsequent processes of GCJproduction.

The ripening (P3) is one of the vital processes in which signifi-cant changes occur in various nutrients and functional propertiesin GCJ (Nisshida et al., 1993). Adding meju powder and subsequentmicrobial fermentation while P3 (ripening) largely affects bio-transformation of proteins, carbohydrates, and glucosides into sim-ple molecules viz., amino acids, sugars, and aglycones (Lee et al.,2012). In particular, amino acids are closely connected to theunique taste, nutrition, and functional aspects of GCJ, and are gen-erally produced by protein degradation (Kang et al., 2011). Alanine,glycine, serine, glutamic acid, and isoleucine are related to sweet,bitter, and savory tastes (Lee, Lee, et al., 2014). Among them, glu-tamic acid is highly correlated with umami taste (Kaneko,Kumazawa, Masuda, Henze, & Hofmann, 2006). In addition, ala-nine, proline, and threonine are known to contribute to the sweettaste in fermented soy-foods including GCJ (Villares, Rostagno,García-Lafuente, Guillamón, & Martínez, 2011). During P3, lysoPCswere significantly decreased, while most of the metabolites viz.,amino acids, organic acids, sugars & sugar alcohols, isoflavones,soyasaponins, and tricin were relatively increased (Fig. 5). Thesemetabolic drifts might have contributed to the enhanced bioactiv-ities including ABTS levels, TPC, TFC as well as the characteristicflavors and taste (umami, sweet, and savory) in GCJ. Further, weobserved the rapid increase in the levels of valine while P3 inWR and WT-GCJ, which might have accelerated the microbialgrowth and production of fermentative enzymes (Baek et al.,2010). In addition, the levels of maltose, glucopyranose, and glu-cose also increased significantly during P3, which reportedly influ-ence the mouthfeel and sweet flavor of GCJ (Kim et al., 2012). Thelevels of organic acid were affected by species of microbial strainsused in the preparation of GCJ, which in turn affects the pH levels,taste, and aroma in end products (Baek et al., 2010). Earlier, Villareset al., 2011 and Lee, Seo, Oh, & Lee, 2014 have reported that duringA. oryzaemediated soybean fermentation, acetyl glycoside and gly-cosides are generally transformed into aglycones, causing changesin isoflavone contents and distribution patterns during ripening. Inour study, both aglycone and glycoside isoflavones were increasedduring ripening. Moreover, the quality indicators like titratableacidity and amino-type nitrogen contents were notably increasedduring P3, which were correlated to the observed alterations infree amino acids levels released through enzymatic hydrolysis ofsoy proteins (Kim, Oh, & Shin, 2008). The addition of red pepperpowder and other ingredients in P4 was followed by the specificelevation in the levels of alkaloids (capsaicin and dihydrocapsaicin)and flavonoids (luteolin 7-O-apiosyl glucoside, luteolin-glucoside,apiin, and apigenin glucosides), which effects the quintessentialpiquant spicy flavor in GCJ (Fig. 5). Additionally, the levels oforganic acids viz., acetic acid, malic acid, and citric acid were shar-ply increased during P4, which greatly influence the flavor proper-ties and shelf life of fermented foods (Fig. 4 and Fig. 5).

In conclusion, we performed a comparative evaluation formetabolite levels, bioactivities, and quality parameters for threeGCJ-types while different processes of its industrial manufacturing.The GCJ samples (WR-GCJ, BR-GCJ, WT-GCJ) were examinedaccording to the RM used (white rice, brown rice, and wheat)and the production processes. We delineated the fermentative bio-process and the cascade of metabolomic alterations in correlationwith quality parameters for industrial GCJ production. The initialfermentation of RM was primarily linked with the production ofmetabolites imparting sweet taste to koji (P1). Further, the

424 Y.K. Jang et al. / Food Chemistry 234 (2017) 416–424

blending withmeju powder (P2) and subsequent ripening (P3) pro-cesses caused the metabolites inception resulting in characteristicsavory and umami taste as well as the functional bioactivities ofcommercial GCJ. Ultimately, adding red pepper powder and severalother additives (P4) caused the production of metabolites influenc-ing its sour and spicy taste. The quality parameters changed dras-tically during the production process, with exception to the pHlevels. Hence, this integrated delineation of the dynamic correla-tions among metabolite levels, bioactivities, and quality parame-ters provides an appropriate model towards standardization ofthe industrial GCJ manufacturing using varying substrates.

Acknowledgements

This work was supported by the National Research Foundationof Korea (NRF) grant funded by the Korea government (MSIP) (No.NRF-2014R1A2A1A11050884) and by the Strategic Initiative forMicrobiomes in Agriculture and Food, Ministry of Agriculture, Foodand Rural Affairs, Republic of Korea (as part of the (multi-ministerial) Genome Technology to Business Translation ProgramNo. 916005-2).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.foodchem.2017.04.154.

References

Baek, J. G., Shim, S. M., Kwon, D. Y., Choi, H. K., Lee, C. H., & Kim, Y. S. (2010).Metabolite profiling of Cheonggukjang, a fermented soybean paste, inoculatedwith various Bacillus strains during fermentation. Bioscience, Biotechnology, andBiochemistry, 74, 1860–1868.

Chang, M. Y., Han, C. Y., Wight, T. N., & Chait, A. (2006). Antioxidants inhibit theability of lysophosphatidylcholine to regulate proteoglycan synthesis.Arteriosclerosis, Thrombosis, and Vascular Biology, 26, 494–500.

Cho, J. Y., Lee, H. J., Shin, H. C., Lee, J. M., Park, K. H., & Moon, J. H. (2013). Behavior offlavonoid glycosides contained in Korean red pepper paste (gochujang) duringfermentation: participation of a b-glucosidase inhibitor. Food Science andBiotechnology, 22, 1–8.

Cvetkovic, B. R., Pezo, L. L., Tasic, T., Šaric, L., Kevrešan, Z., & Mastilovic, J. (2015). Theoptimisation of traditional fermentation process of white cabbage (in relation tobiogenic amines and polyamines content and microbiological profile). FoodChemistry, 168, 471–477.

Ghosh, K., Ray, M., Adak, A., Dey, P., Halder, S. K., Das, A., & Mondal, K. C. (2015).Microbial, saccharifying and antioxidant properties of an Indian rice basedfermented beverage. Food Chemistry, 168, 196–202.

Hong, S. W., Heo, H., Yang, J. H., Han, M., Kim, D. H., & Kwon, Y. K. (2013).Soyasaponin I improved neuroprotection and regeneration in memory deficientmodel rats. PLoS One, 8, e81556.

Hong, J. H., Lee, K. W., Chung, S., Chung, L., Kim, H. R., & Kim, K. O. (2012). Sensorycharacteristics and cross-cultural comparisons of consumer acceptability forGochujang dressing. Food Science and Biotechnology, 21, 829–837.

Itoh, H., Kawashima, K., & Chibata, I. (1975). Antioxidant activity of browningproducts of triose sugar and amino acid. Agricultural and Biological Chemistry, 39,283–284.

Kaneko, S., Kumazawa, K., Masuda, H., Henze, A., & Hofmann, T. (2006). Molecularand sensory studies on the umami taste of Japanese green tea. Journal ofAgricultural and Food Chemistry, 54, 2688–2694.

Kang, H. J., Yang, H. J., Kim, M. J., Han, E. S., Kim, H. J., & Kwon, D. Y. (2011).Metabolomic analysis of meju during fermentation by ultra-performance liquidchromatography-quadrupole-time of flight mass spectrometry (UPLC-Q-TOFMS). Food Chemistry, 127, 1056–1064.

Kim, A. J., Choi, J. N., Kim, J., Kim, H. Y., Park, S. B., Yeo, S. H., et al. (2012). Metaboliteprofiling and bioactivity of rice koji fermented by Aspergillus strains. Journal ofMicrobiology and Biotechnology, 22, 100–106.

Kim, H. E., Han, S. Y., & Kim, Y. S. (2010). Quality characteristics of gochujang mejuprepared with different fermentation tools and inoculation time of Aspergillusoryzae. Food Science and Biotechnology, 19, 1579–1585.

Kim, Y. S., Oh, B. H., & Shin, D. H. (2008). Quality characteristics of kochujangprepared with different meju fermented with Aspergillus sp. and Bacillus subtilis.Food Science and Biotechnology, 17, 527–533.

Kim, J. Y., Park, K. W., Yang, H. S., Cho, Y. S., Jeong, C. H., Shim, K. H., et al. (2005).Anticancer and immuno-activity of methanol extract from onion Kochujang.Korean Journal of Food Preservation, 12, 173–178.

Kim, J. Y., & Yi, Y. H. (2010). pH, acidity, color, amino acids, reducing sugars, totalsugars, and alcohol in puffed millet powder containing millet Takju duringfermentation. Korean Journal of Food Science and Technology, 42, 727–732.

Korea Agro-Fisheries & Food Trade Corporation (2015). Production and import &export of processed foods: Gochujang..

Lee, S. Y., Kim, H. Y., Lee, S., Lee, J. M., Muthaiya, M. J., Kim, B. S., et al. (2012). Massspectrometry-based metabolite profiling and bacterial diversitycharacterization of Korean traditional meju during fermentation. Journal ofmicrobiology and biotechnology, 22, 1523–1531.

Lee, J. S., Kim, D. H., Liu, K. H., Oh, T. K., & Lee, C. H. (2005). Identification offlavonoids using liquid chromatography with electrospray ionization and iontrap tandem mass spectrometry with an MS/MS library. Rapid Communicationsin Mass Spectrometry, 19, 3539–3548.

Lee, D. E., Lee, S., Jang, E. S., Shin, H. W., Moon, B. S., & Lee, C. H. (2016). Metabolomicprofiles of aspergillus oryzae and bacillus amyloliquefaciens during rice Kojifermentation. Molecules, 21(6), 773.

Lee, S. Y., Lee, S., Lee, S., Oh, J. Y., Jeon, E. J., Ryu, H. S., et al. (2014). Primary andsecondary metabolite profiling of doenjang, a fermented soybean paste duringindustrial processing. Food Chemistry, 165, 157–166.

Lee, S., Seo, M. H., Oh, D. K., & Lee, C. H. (2014). Targeted metabolomics forAspergillus oryzae-mediated biotransformation of soybean isoflavones, showingvariations in primary metabolites. Bioscience, Biotechnology, and Biochemistry,78, 167–174.

Lin, Y., Meijer, G. W., Vermeer, M. A., & Trautwein, E. A. (2004). Soy protein enhancesthe cholesterol-lowering effect of plant sterol esters in cholesterol-fedhamsters. The Journal of nutrition, 134, 143–148.

Moongngarm, A., & Saetung, N. (2010). Comparison of chemical compositions andbioactive compounds of germinated rough rice and brown rice. Food Chemistry,122, 782–788.

Nisshida, K., Ohta, Y., Araki, Y., Ito, M., Nagamura, Y., & Ishiguro, I. (1993). Inhibitoryeffects of ‘‘Group A Saponin”, and ‘‘Group B Saponin” fractions from soybeanseed hypocotyls on radical-initiated lipid peroxidation in mouse livermicrosomes. Journal of Clinical Biochemistry and Nutrition, 15, 175–184.

Nout, M. J. R. (1994). Fermented foods and food safety. Food Research International,27, 291–298.

Oh, D. G., Jang, Y. K., Woo, J. E., Kim, J. S., & Lee, C. H. (2016). Metabolomics revealsthe effect of garlic on antioxidant-and protease-activities during Cheonggukjang(fermented soybean paste) fermentation. Food Research International, 82, 86–94.

Park, K. Y., Kong, K. R., Jung, K. O., & Rhee, S. H. (2001). Inhibitory effects ofkochujang extracts on the tumor formation and lung metastasis in mice.Preventive Nutrition and Food Science, 63, 187–191.

Salmond, C. V., Kroll, R. G., & Booth, I. R. (1984). The effect of food preservatives onpH homeostasis in Escherichia coli. Microbiology, 130(11), 2845–2850.

Shin, H. W., Jang, E. S., Moon, B. S., Lee, J. J., Lee, D. E., Lee, C. H., et al. (2016). Anti-obesity effects of gochujang products prepared using rice koji and soybean mejuin rats. Journal of Food Science and Technology, 53, 1004–1013.

Shin, D. H., Kim, D. H., Choi, U., Lim, D. K., & Lim, M. S. (1996). Studies on thephysicochemical characteristics of traditional kochujang. Korean Journal of FoodScience and Technology, 28, 157–161.

Suh, D. H., Jung, E. S., Park, H. M., Kim, S. H., Lee, S., Jo, Y. H., et al. (2016). Comparisonof metabolites variation and antiobesity effects of fermented versusnonfermented mixtures of Cudrania tricuspidata, Lonicera caerulea, andsoybean according to fermentation in vitro and in vivo. PLoS One, 11, e0149022.

Villares, A., Rostagno, M. A., García-Lafuente, A., Guillamón, E., & Martínez, J. A.(2011). Content and profile of isoflavones in soy-based foods as a function of theproduction process. Food and Bioprocess Technology, 4, 27–38.

Yang, X., Dong, C., & Ren, G. (2011). Effect of soyasaponins-rich extract fromsoybean on acute alcohol-induced hepatotoxicity in mice. Journal of Agriculturaland Food Chemistry, 59, 1138–1144.

Yu, K. W., Lee, S. E., Choi, H. S., Suh, H. J., Ra, K. S., Choi, J. W., et al. (2012).Optimization for rice koji preparation using Aspergillus oryzae CJCM-4 isolatedfrom a Korean traditional meju. Food Science and Biotechnology, 21, 129–135.

Zhou, B., Xiao, J. F., Tuli, L., & Ressom, H. W. (2012). LC-MS-based metabolomics.Molecular BioSystems, 8, 470–481.