Effect of fermentation stages on glucosinolate profiles in kimchi: Quantification of 14 intact glucosinolates using ultra-performance liquid chromatography-tandem mass spectrometry

Highlights • An analytical method for estimating glucosinolate profiles of kimchi is reported.• The method employs ultra-performance liquid chromatography-tandem mass spectrometry.• The method is efficient in terms of linearity, sensitivity, accuracy, and precision.• The glucosinolate contents and compositions vary with fermentation stage.• Total glucosinolates were degraded by 91%–100% in over-fermentation stage.


Introduction
Regular consumption of cruciferous vegetables has been proposed to benefit human health, including cancer prevention and antiinflammatory effects, mainly attributed to glucosinolate-derived isothiocyanates and indoles (Fuentes et al., 2015;Esteve, 2020). Glucosinolates are sulfur-containing secondary metabolites abundant in cruciferous vegetables, such as broccoli, cabbage, and other green leafy vegetables and are responsible for their characteristic smell and taste. These compounds are biologically inert; however, they can be degraded into isothiocyanates, nitriles, epithionitriles, thiocyanates, and indoles, based on the type of glucosinolate, environmental pH, and the presence of specific proteins. The degradation is caused by coexisting myrosinase, a thioglucosidase, activated upon maceration of tissues (Hanschen et al., 2014). Intact glucosinolates can also partly be metabolized to form breakdown products such as isothiocyanates by the myrosinase-like activity of the human gut microbiota (Shakour et al., 2022).
Kimchi, listed in the Codex Alimentarius in 2001 (CODEX STAN 223-2001), is a globally known traditional fermented food. Kimchi constitutes a major component of Korean food and is composed of fermented vegetables, primarily brined cabbage (Brassica rapa L. subsp. pekinensis) and mixed with various seasonings such as red pepper (Capsicum annuum L.) powder, garlic, ginger, and edible Allium varieties. In Korea, kimchi is frequently consumed in substantial amounts (the average daily intake of kimchi was approximately 64 g in 2015-2019). Therefore, it is likely to be the primary dietary source of glucosinolates and their breakdown products in Korea. Moreover, Korean kimchi exports have increased drastically in recent years, indicating its increased global consumption and popularity. Therefore, to realize the therapeutic potentials of kimchi, investigation and estimation of the glucosinolates and their breakdown products in kimchi is important.
Total glucosinolate content in kimchi cabbage varies in the range of 2.70-57.88 µmol/g dry weight (DW) based on the variety, and gluconapin, glucobrassicanapin, and 4-methoxyglucobrassicin are the major glucosinolates (Baek et al., 2016;Kim et al., 2010;Lee et al., 2014;Chun et al., 2018). Glucosinolates in kimchi cabbage can be partly lost because of their interaction with myrosinase or leaching during kimchi preparation, which involves trimming, cutting, salting, and seasoning, thus damaging plant tissues. Moreover, during kimchi fermentation, glucosinolates can be further degraded into breakdown products by myrosinase and myrosinase-like bacterial enzymes (Hanschen et al., 2014). Collectively, various factors, including processing and fermentation conditions and the intrinsic quality of kimchi cabbage, which is determined by the variety, growth, and storage conditions, are the major determinants of the glucosinolate content of kimchi. However, the glucosinolate profile of kimchi and the factors affecting it have garnered less attention. To date, only one study has reported relative quantities of glucosinolates in kimchi products (Kim et al., 2017). This study identified glucoalyssin (0.00-7.07 µmol/g DW), gluconapin (0.00-5.85 µmol/ g DW), glucobrassicanapin (0.00-11.87 µmol/g DW), glucobrassicin (0.00-0.42 µmol/g DW), and 4-methoxyglucobrassicin (0.12-9.36 µmol/g DW) in kimchi samples; however, it did not consider the effects of various fermentation processes on the glucosinolate contents.
Glucosinolates are typically determined by analyzing desulfoglucosinolates after the enzymatic desulfation of intact glucosinolates using reversed-phase liquid chromatography coupled with ultraviolet or diode array detection (Hennig et al., 2012;Klopsch et al., 2017). However, with the advancement of technologies, electrospray ionization mass spectrometry (ESI-MS), which avoids the time-consuming and poorly controlled desulfation step (Bernal et al., 2019;Hooshmand & Fomsgaard, 2021), has been used for the profiling of intact glucosinolates. Moreover, ultra-performance liquid chromatography (UPLC) and ultrahigh-performance liquid chromatography have been shown to improve the resolution and sensitivity of the techniques and enable faster separation of glucosinolates (Thomas et al., 2018;Capriotti et al., 2018;Bernal et al., 2019). The complexity of the kimchi matrix comprising various ingredients (Codex Alimentarius Commission, 2001), makes it difficult to obtain reliable and accurate results. Therefore, determining glucosinolates in kimchi requires an efficient clean-up process to eliminate substances that interfere with this analysis.
We hypothesized that developing and validating an efficient analytical method using UPLC-ESI-tandem mass spectrometry (MS/MS) can efficiently determine intact glucosinolates in kimchi. To test this hypothesis, we aimed to develop and validate a method based on UPLC-ESI-MS/MS and assess the effects of fermentation stages on the glucosinolate profiles of kimchi. To the best of our knowledge, this study is the first to determine the effect of fermentation stage on glucosinolate profiles in kimchi using a validated analytical method for quantification of glucosinolates in intact forms. This study is expected to contribute to our understanding of changes in glucosinolates during kimchi fermentation and improve the efficiency of glucosinolate quantification in kimchi.

Sample preparation
Twenty kimchi products freshly prepared from kimchi cabbage were purchased from various manufacturers in Korea. Titratable acidities (as lactic acid) of the filtrates of homogenized kimchi samples (10 mL) were measured by titrating with 0.1 N sodium hydroxide solution to pH 8.3 using an automatic titrator (Model TitroLine 5000; SI Analytics, Mainz, Germany). These kimchi samples had titratable acidity of <0.5 % and were classified as non-fermented kimchi. To obtain moderate-fermented and over-fermented kimchi, the collected kimchi samples were stored at 6 • C until titratable acidities were either ≥0.6 and ≤1.0 (moderatefermented kimchi), which took 1-2 weeks; or >1.0 % (over-fermented kimchi), which took more than 3 weeks.

Sample treatment
Approximately 50 mg of each freeze-dried and ground sample was transferred into a 15 mL conical tube with a cap and mixed with 10 mL 70 % (v/v) methanol. The mixture was extracted by sonication for 10 min at room temperature and then centrifuged at 3,900 rpm for 20 min at 4 • C. After removing methanol from the collected supernatant using a rotary evaporator, the volume of the remaining aqueous phase was adjusted to 10 mL with deionized water. The resulting solution was filtered through a 0.20 μm syringe filter. For clean-up, 2 mL of the filtrate was loaded onto a weak anion-exchange solid-phase extraction (SPE) cartridge (Oasis WAX 3 cc cartridge, weight: 60 mg, particle size: 30 µm, Waters) that was previously conditioned with 3 mL methanol and activated with 3 mL 2 % (v/v) formic acid. The cartridge was sequentially washed with 1 mL 2 % formic acid and 1 mL methanol. After drying for 2 min under vacuum, the analytes were eluted with 5 % (v/v) ammonia solution (≥28.0 %) in methanol (10 mL). The eluted solution was thoroughly dried by rotary evaporation and reconstituted with 2 mL deionized water. The resulting solution was used for UPLC-ESI-MS/MS analysis.

UPLC-ESI-MS/MS analysis
Intact glucosinolates were quantified using an Acquity UPLC® I-Class system coupled with a tandem quadrupole mass spectrometer (Xevo TQ-S) equipped with an ESI source. Chromatographic separation was performed using an Acquity UPLC® BEH C18 column (2.1 × 150 mm, 1.7 µm, Waters) with a mixture of 0.1 % (v/v) formic acid in water (A) and 0.1 % (v/v) formic acid in acetonitrile (B) as the mobile phase using the following linear gradient conditions: 100 % (v/v) of A for 0-1 min, 100 % to 95 % of A for 1-3 min, 95 % to 70 % of A for 3-6.2 min, 70 % of A for 6.2-7.2 min, 70 % to 0 % of A for 7.2-8 min, 0 % of A for 8-9 min, 0 % to 100 % of A for 9-10 min, and 100 % of A for 10-12 min. The column temperature was maintained at 30 • C, the flow rate was 0.25 mL/min, and the injection volume was 1 µL. ESI was performed with the negative-ion mode under the following conditions: capillary voltage, 3 kV; desolvation temperature, 350 • C; desolvation gas flow, 650 L/h; cone gas flow, 150 L/h; source temperature, 150 • C.

Method validation
Method validation was performed according to international guidelines (Association of Analytical Communities (AOAC), 2012; United States Food and Drug Administration (USFDA), 2019). To determine the selectivity of the proposed method, the chromatograms of the samples were compared with those of standard solutions and samples spiked with standard solutions. The limit of detection (LOD) and quantification (LOQ) values were calculated using the following formula: where σ and S are the standard deviation of the y-intercept and the slope obtained from triplicate calibration curves with 5-8 points for each analyte, respectively. The calibration ranges were as follows: 7-1000 nmol/L for sinigrin, gluconapin, glucobrassicanapin, progoitrin, glucoiberin, gluconasturtiin, glucoberteroin, glucoraphanin, glucocheirolin, glucobrassicin, and glucoalyssin; 10-1000 nmol/L for glucoraphenin; 7-3000 nmol/L for 4-methoxyglucobrassicin; and 3-1000 nmol/L for neoglucobrassicin. The matrix effect was evaluated by comparing the analyte concentrations in the blank matrix (sample dissolution solvent; deionized water) spiked with standard solutions at three concentration levels (low, 100 nmol/L; medium, 300 nmol/L; and high, 500 nmol/L) to those in the sample matrix spiked with standard solutions of the same concentrations after extraction. The matrix effect (n = 6) was calculated using the following formula: where A is the analyte concentration in the sample matrix spiked after extraction, B is the analyte concentration in the non-spiked sample matrix, and C is the analyte concentration in the spiked blank matrix (sample dissolution solvent; deionized water).
The accuracy and the intraday and interday precision were evaluated by comparing the analyte concentrations in the samples spiked with standard solutions of three concentrations (low, 100 nmol/L; medium, 300 nmol/L; and high, 500 nmol/L) after extraction to those in the samples spiked with standard solutions of the same concentrations before extraction. The accuracy and intraday precision data were collected on the same day (n = 6), whereas the interday precision data were collected on three consecutive days (n = 6). The accuracy was expressed as a percentage of recovery using the following formula: where A is the analyte concentration in the sample matrix spiked before extraction, B is the analyte concentration in the sample matrix spiked after extraction, and C is the analyte concentration in the nonspiked sample matrix. Precision was expressed as relative standard deviation (RSD) values.

Statistical analysis
Means and standard deviations of data were calculated using IBM SPSS Version 19.

UPLC
Several experiments were performed to establish efficient gradient elution conditions using a mixture of standard solutions, different mobile phase compositions, and flow rate conditions. The best results were obtained using the mobile phase gradient described in section 2.4. As shown in Figure S2, the chromatographic separation of 14 intact glucosinolates was completed within 9 min. The overall run time required to obtain a reproducible retention time was 12 min, which is shorter than that reported previously (Thomas et al., 2018;Bernal et al., 2019;Hooshmand & Fomsgaard, 2021).

ESI-MS/MS
Optimal multiple reaction monitoring parameters for the 14 intact glucosinolates were established and are summarized in Table S1. The optimal operating parameters (i.e., cone voltage and collision energy) for the two most intense transitions (one precursor ion → two product ions) for each intact glucosinolate were determined by directly injecting each standard solution into the mass spectrometer operated in negative ionization mode. The most abundant product ion formed from each intense precursor [M− H] − occurred at m/z 97, corresponding to the sulfate moiety of glucosinolates, and was selected to quantify the intact glucosinolates. The second most abundant product ion was monitored together with the product ion at m/z 97 to identify intact glucosinolates. The ions for confirmation of analytes are summarized in Table S1. The precursor and product ions selected in this study for intact glucosinolate analysis are commonly used in an MS/MS system (Thomas et al., 2018;Capriotti et al., 2018).

Sample extraction and clean-up treatment
Intact glucosinolates were extracted by the ultrasonic extraction method using 70 % (v/v) methanol as the extraction solvent. To determine the appropriate extraction time, the extractions were performed for 10, 30, and 60 min. The optimal extraction time was found to be 10 min, as no remarkable improvements in the extraction efficiency were observed when longer extraction times were used.
A weak anion-exchange SPE cartridge was used for the extraction clean-up process. Several solvents were evaluated to establish an effective SPE procedure. For the washing step, combinations of 1 or 2 mL 2 % (v/v) formic acid in water followed by 0.5 or 1 mL methanol were tested. The most effective solvents were 1 mL 2 % formic acid in water, followed by 1 mL methanol. For the elution step, suitable recoveries (80 %-110 %) were obtained using 10 mL 5 % (v/v) ammonia solution (≥28.0 %) in methanol. The established SPE treatment was effective in eliminating interference that affected the matrix effect (see section 3.4.2).

Selectivity
To evaluate the selectivity of the proposed method, the chromatograms of the standard solutions were compared with those of kimchi samples spiked with a mixture of standard solutions, as obtaining a glucosinolate-free kimchi sample as a blank matrix was impossible ( Figure S2). No interfering peaks were observed at the retention time for any intact glucosinolates, indicating the absence of interference from coexisting matrix components. These results indicated that the proposed method is selective for determining intact glucosinolates in kimchi.

Matrix effect
To evaluate the effect of the kimchi matrix on the ESI process, we assessed the matrix effect in the kimchi extract, and significant ion suppression or enhancement was observed for several intact glucosinolates when the clean-up treatment was not performed (Table S2). These results indicated that co-eluting compounds that change the ionization efficiencies and affect the quantification of intact glucosinolates might be present in the kimchi matrix. In addition, sample clean-up is required to eliminate interference, even when using MS/MS, one of the most sensitive and selective detection systems, because of the unavoidability of accompanying matrix effects during the analysis of complicated matrices using ESI (Matuszewski et al., 2003;Zhou et al., 2017). Therefore, we attempted to improve the matrix effect using SPE treatment. This strategy seemed more suitable than reducing the injection volume, diluting the sample, or using matrix-matched calibration. Furthermore, the concentrations of some glucosinolates in the kimchi matrix might not be high enough to facilitate quantification after dilution. Moreover, the relative expense of the standard solutions makes the matrix-matched calibration expensive for the analysis of large numbers of samples.
The use of SPE treatment successfully eliminated unintended interference in the kimchi matrix. It improved the matrix effects to 98 %-105 % for the intact glucosinolates at the three spiked concentrations, except for that with 4-methoxyglucobrassicin, for which the matrix effect was 88 %-91 % (Table 1). These results confirmed the absence of interference that might considerably alter the ESI process, indicating that the proposed method is selective for determining intact glucosinolates in the kimchi matrix.

Linearity and sensitivity
The calibration curves of all intact glucosinolates showed excellent coefficients of determination (r 2 ≥ 0.9991; Table 2). Moreover, the LOD and LOQ values of all intact glucosinolates were lower than 11 and 35 nmol/L, respectively, which were lower than those reported previously (Bernal et al., 2019;Hooshmand & Fomsgaard, 2021;Maldini et al., 2017). Table 3 summarizes the accuracy results, as evaluated for the recovery of intact glucosinolates from the kimchi matrix. The recovery of intact glucosinolates was in the range of 83 %-92 % at the low concentration, 82 %-95 % at the medium concentration, and 86 %-101 % at the high concentration. The recovery of all intact glucosinolates was within the range of 80 %-110 %, which is considered acceptable according to international guidelines (AOAC International, 2012; USFDA, 2019).

Intraday and interday precision
The intraday and interday RSD values for the 14 intact glucosinolates at the three spiked concentrations were used to assess the precision of the proposed method (Table 4). The intraday RSD values ranged from 3 % to 8 % at the low concentration, 3 % to 8 % at the medium concentration, and 2 % to 7 % at the high concentration. The interday RSD values were in the range of 5 %-8% at the low concentration, 4 %-8% at the medium concentration level, and 2 %-7% at the high concentration.