Determination of the Metabolite Content of Brassica juncea Cultivars Using Comprehensive Two-Dimensional Liquid Chromatography Coupled with a Photodiode Array and Mass Spectrometry Detection

Plant-based foods are characterized by significant amounts of bioactive molecules with desirable health benefits beyond basic nutrition. The Brassicaceae (Cruciferae) family consists of 350 genera; among them, Brassica is the most important one, which includes some crops and species of great worldwide economic importance. In this work, the metabolite content of three different cultivars of Brassica juncea, namely ISCI Top, “Broad-leaf,” and ISCI 99, was determined using comprehensive two-dimensional liquid chromatography coupled with a photodiode array and mass spectrometry detection. The analyses were carried out under reversed-phase conditions in both dimensions, using a combination of a 250-mm microbore cyano column and a 50-mm RP-Amide column in the first and second dimension (2D), respectively. A multi (three-step) segmented-in-fraction gradient for the 2D separation was advantageously investigated here for the first time, leading to the identification of 37 metabolites. In terms of resolving power, orthogonality values ranged from 62% to 69%, whereas the corrected peak capacity values were the highest for B. juncea ISCI Top (639), followed by B. juncea “Broad-leaf” (502). Regarding quantification, B. juncea cv. “Broad-leaf” presented the highest flavonoid content (1962.61 mg/kg) followed by B. juncea cv. ISCI Top (1002.03 mg/kg) and B. juncea cv. ISCI 99 (211.37 mg/kg).


Introduction
Vegetables from the Brassicaceae or Cruciferae family represent the most commonly consumed vegetables worldwide. This family includes brussels sprouts, broccoli, cabbage, cauliflower, and others. Such vegetables do contain high levels of bioactive compounds, e.g., polyphenols, carotenoids, tocopherols, glucosinolates, and ascorbic acid [1][2][3][4]. Epidemiological data have demonstrated the

Results and Discussion
The analysis of the three different cultivars of B. juncea L. was first run using a conventional LC-PDA-MS approach on a C18 column. As illustrated in the following section, a considerable number of compounds overlapped; consequently, an RP-LC×RP-LC system was adopted in order to attain higher separation power, thus providing a thorough overview of the overall metabolites pool, which is beneficial for quantification purposes.

Elucidation of Brassica juncea Cultivars Using RP-LC×RP-LC-PDA-MS
RP-LC×RP-LC separations have proved to be quite effective for the analysis of the metabolite content of food and natural products [19][20][21][22]25,[30][31][32]. Before running an RP-LC×RP-LC analysis, a careful optimization of the independent separations must be carried out [26,27,29]. A low mobile phase flow rate is preferred in the 1 D in order to decrease the fraction volume onto the 2 D and augment the 1 D sampling rate. Usually, this is achieved by employing a microcolumn in the 1 D; however, since most commercial LC pumps are not capable of delivering a stable and repeatable flow rate, a higher flow rate is commonly employed and split up before entering the 1 D column. A scheme of the RP-LC×RP-LC employed is reported in Figure 1. In this work, a robust and easy-to-use micropump with a completely new direct-drive engineering was advantageously employed, and was capable of delivering micro-to semi-micro flow rates ranging from 1 to 500 µL/min. Repeatability data obtained on four selected peaks are displayed in Table 1. Relative standard deviation (RSD, %) values lower than 0.02 were attained in the case of mean retention times (min), whereas RSD (%) values lower than 1.21 were determined in case of mean areas.  With regard to the 2 D, a fast separation is commonly employed in order to increase the 1 D sampling to lower the risk of incurring wrap-around phenomena. Consequently, a microcyano column was chosen in the 1 D, whereas a 4.6-mm I.D. partially porous RP-Amide column was employed in the 2 D and operated at 2 mL min −1 . For fraction transfer, two high-speed, six-port, two-position switching valves equipped with two 10 µL sampling loops were chosen.
In this context, the optimization of the gradient programs, especially for the 2 D, is also necessary for an adequate separation and is mainly related to the chemical properties of the solutes. Late eluting compounds that are retained more in the 2 D require a greater gradient steepness in order not to incur wrap-around effects. In the case of closely related compounds, e.g., early-eluting compounds, which are subjected to co-elutions, a lower gradient of steepness is preferable in order to permit stronger retention.
Following this strategy, a newly developed RP-LC×RP-LC system was investigated. In particular, a multi segmented-in-fraction gradient approach was employed, as illustrated in Figure 2. In particular, three different full-in-fraction gradients were considered for the 2 D analysis. The first gradient was from 10 to 32 min, where %B ranged from 3% to 8% (∆%B: 5) for the analysis of early eluting organic acids; in the second gradient step (from 32 to 43 min), %B ranged from 10% to 44% (∆%B: 34) for the analysis of (acetylated) tri-and tetrasaccharides, whereas in the last one (from 43 to 60 min), %B ranged from 20% to 60% (∆%B: 40) for the analysis of late eluting (acetylated) mono-and disaccharides. The modulation time of the switching valves was 1.00 min. Figure 2 shows the contour plots for the RP-LC×RP-LC analysis of the three cultivars of Brassica juncea, where a total of 37, 34, and 31 metabolites were positively separated using the optimized multi segmented-in-fraction gradient approach.
Concerning the performance of the developed RP-LC×RP-LC system, Table 2 reports the values attained for both peak capacity and orthogonality [33]. The highest theoretical peak capacity values, which are multiplicative of the peak capacity of the two single dimensions [34], were attained for the cultivar ISCI Top (1734), whereas the lowest was obtained for the cultivar ISCI 99 (932). The orthogonality values ranged from 62% to 69% for ISCI Top and "Broad leaf", respectively [33]. The corrected peak capacity values, which considered, both undersampling [35] and orthogonality values, were 639, 404, and 502 for ISCI Top, ISCI 99, and "Broad leaf", respectively. Considering the similarity of the two separation systems employed in both dimensions, such values can be considered quite remarkable and are in agreement with previously published findings on similar set-ups exploited for polyphenolic characterization in licorice (695 in Wong et al. [30]) and pistachio (461-633 in Arena et al. [31]) samples. As an example, the benefits associated with the employment of the developed RP-LC×RP-LC with the multi segmented-in-fraction gradient program over the conventional RP-LC separation are highlighted in Figure 3.
A selected chromatographic region of the Brassica ISCI Top extract ( Figure 3A) clearly shows how the 1D-LC did not provide enough peak capacity for unambiguous characterization of the chemical profile of the three occurring metabolites, due to compound overlapping. However, when the RP-LC×RP-LC analysis was employed, the three different bioactive compounds were conveniently separated and characterized via inspection of the respective MS spectra ( Figure 3B). As a result, the better resolution of the RP-LC×RP-LC separation (with the 2 D operated under the multi (three-step) segmented-in-fraction gradient mode) over the conventional 1D-LC led to a greater metabolite expansion in the RP-LC×RP-LC space, which was essential for improving the reliable identification of compounds with complexity and/or various polarities.

Semi-Quantitative Determination of the Flavonoid Content of Brassica juncea Cultivars
Tentative identification of the Brassica juncea extracts, illustrated in Figure 2, was performed based on their PDA, MS, and literature data [1,2,[9][10][11][14][15][16][36][37][38]. Among the major classes of compounds identified, organic acids, (acetylated) tri-and tetrasaccharides, and (acetylated) mono-and disaccharides, were recognized (Table 3). Due to the lack of commercial standards, quantification of Brassica spp content has so far been carried out after acidic and/or alkaline hydrolysis [36][37][38]. In this work, a quantification of the native flavonoid composition of the three cultivars of Brassica juncea was carried out by RP-LC×RP-LC system coupled to PDA detection for the first time. Toward such an aim, and considering the unavailability of corresponding standard references, an established approach in the field of food and natural product analysis was followed. Basically, three standards, as representatives of the distinct chemical classes, i.e., Km 3-O-glucoside, Isorhamnetin (Is) 3-O-glucoside, and Qn 3-O-glucopyranoside, were chosen and calibration curves were prepared, as reported in Section 3.4.5. Results are shown in Table 4, which reports all the standard curves, correlation coefficients (R 2 ), limits of detection (LoDs) and limits of quantification (LoQs), and relative standard deviations (RSDs) of the peak areas for each standard selected. The five-point calibration curves provided R 2 values ranging from 0.9993 to 0.9997, whereas for LoQ and LoD, values as low as only 30 ppb and 90 ppb, respectively, were found. Finally, RSD values lower than 0.89% were obtained, demonstrating valuable method repeatability.   Subsequently, all three samples were analyzed and the contents of the target compounds were calculated using commercially-available software, as reported in Table 3   1D-LC separations were performed on an Ascentis Express C18 column (Merck Life Science, Merck KGaA, Darmstadt, Germany; 150 × 4.6 mm I.D., 2.7 µm dp). LC×LC separations were conducted by using a 1 D Ascentis ES-Cyano (ES-CN) column (Merck Life Science, Merck KGaA, Darmstadt, Germany; 250 × 1.0 mm I.D., 5 µm dp) and a 2 D Ascentis Express RP-Amide column (Merck Life Science, Merck KGaA, Darmstadt, Germany; 50 × 4.6 mm I.D., 2.7 µm dp).

Sample and Sample Preparation
Brassica juncea L. Czern & Coss cv. ISCI 99, ISCI Top, and "Broad-leaf" leaf selections were provided from the Brassica collection of Consiglio per la ricerca in agricoltura e l'analisi dell'economia agraria -Centro di Ricerca Cerealicoltura e Colture Industriali) (CREA-CI) [39]. Samples were immediately frozen and freeze-dried for storage in glass vacuum desiccators. Lyophilized tissues were finely powdered to 0.5 µm size for analysis. Compound extraction was carried out based on the following protocol [16] with some modifications. The powder of the leaves of the three different B. juncea cultivars were weighed into 100 mg samples. The samples were extracted twice with 5 mL of a mixture of methanol:water (60:40, v/v) for 30 min in a sonicator and centrifuged at 1000× g for 15 min, followed by filtration of the supernatants through a 0.45-µm nylon filter (Merck Life Science, Merck KGaA, Darmstadt, Germany). The prepared organic extracts were subjected to evaporation in a EZ-2 evaporator and then redissolved in 1 mL of the same solvent extraction mixture of methanol:water (60:40, v/v).

Instrumentation
LC and LC × LC analyses were performed on a Nexera-e liquid chromatograph (Shimadzu, Kyoto, Japan), consisting of a CBM-20A controller, one LC-Mikros binary pump, two LC-30AD dual-plunger parallel-flow pumps, a DGU-20A 5 R degasser, a CTO-20AC column oven, a SIL-30AC autosampler, and an SPD-M30A PDA detector (1.0 µL detector flow cell volume). The two dimensions were connected by means of two high-speed/high-pressure, two-position, six-port switching valves with a micro-electric actuator (model FCV-32 AH, 1.034 bar; Shimadzu, Kyoto, Japan), placed inside the column oven, and equipped with two 10-µL stainless steel loops. The Nexera-e liquid chromatograph was hyphenated to an LCMS-8050 triple quad mass spectrometer through an ESI source (Shimadzu, Kyoto, Japan).

Data Handling
The LC × LC-LCMS-8050 system and the switching valves were controlled using the Shimadzu Labsolution software (ver. 5.93) (Kyoto, Japan). LC×LC-Assist software (ver. 2.00) (Shimadzu, Kyoto, Japan) was used for setting up the multi (three-step) segmented-in-fraction gradient analyses. The LC × LC data were visualized and elaborated into two and three dimensions using Chromsquare ver. 2.3 software (Shimadzu, Kyoto, Japan).

Construction of Calibration Curves
For flavonoid determination, due to the lack of commercial standards, Km 3-O-glucoside, Is 3-O-glucoside, and Qn 3-O-glucopyranoside, as representatives of the distinct chemical classes under evaluation, were selected. Standard calibration curves were prepared in the concentration range 0.1-100 mg L −1 with five different concentration levels, run in triplicate. The amount of the compound was finally expressed in mg kg −1 of extract.

Conclusions
In this paper, the benefits associated with the use of a multi (three-step) segmented-in-fraction gradient in the RP-LC×RP-LC-PDA-MS analysis of three Brassica juncea cultivars are demonstrated. The coupling of a microcyano and an RP-Amide columns, in the first and second dimension, respectively, provided a characteristic metabolite pattern of the extracts, leading to the identification of 37 bioactives of different chemical nature, i.e., organic acids, (acetylated) tri-and tetrasaccharides, and (acetylated) mono-and disaccharides. Interestingly, the employment of a micro LC pump in the first dimension of the RP-LC×RP-LC-PDA-MS systems allowed for high repeatability and stable retention times and areas. The investigated approach can be advantageously employed for RP-LC×RP-LC metabolic analyses of other complex plant derived extracts.