Multielement Analysis of Pakchoi (Brassica rapa L. ssp. chinensis) by ICP-MS and Their Classification according to Different Small Geographical Origins

Statistical interpretation of the concentrations of 42 elements, determined using solution-based inductively coupled plasma mass spectrometry (ICP-MS) analysis and multivariate statistical methods, such as principal components analysis (PCA), was used to establish the provenance of pakchoi (Brassica rapa L. ssp. chinensis) from 6 areas in Ha Noi, Vietnam. Although pakchoi is widely cultivated and manufactured, no universal method is used to discriminate the origin of this vegetable. Our study introduced for the first time a method to classify pakchoi in small geographical areas. 42 metallic elements of pakchoi were detected by ICP-MS, which were further analyzed using multivariate statistical analysis to perform clusters based on the geographical locations. Eleven elements, i.e., 28Si; 56Fe; 59Co; 63Cu; 69Ga; 75As; 85Rb; 93Nb; 107Ag; 118Sn, and 137Ba, were identified as discriminators to distinguish pakchoi from those areas. Results from this study presented a new method to discriminant the geographical origins of pakchoi, which could apply to other types of vegetables on the food market.


Introduction
Discriminating the geographical origins of various food and drink products is growing considerably in recent years. In general, inductively coupled plasma mass spectrometry (ICP-MS) is one of the most useful approaches for multielement analysis of foods and drinks. It has been shown that ICP-MS data could be used successfully for discriminating wines from different geographical areas in Slovenia, Romania, and Croatia [1][2][3]. Chuzinska and Baralkiewicz determined 13 mineral elements for the classification of 55 honey samples in Poland [4]. Whereas, Chung et al. distinguished verified cultivation regions of potato by the stable isotope composition analysis of bioelements [5]. e geographical origin of vegetables, such as cabbages, was also discriminated by 21 inorganic elements through ICP-MS analysis [6]. e data of elements compositions are often combined with multivariate statistical analysis, such as principal component analysis (PCA), to reduce the dimension of the data matrix, visualize the similarities and differences among samples, and identify the key discriminants distinguishing the samples [7].
Pakchoi (Brassica rapa L.) is a major vegetable in Vietnamese diets. e consumption of pakchoi may provide many health benefits. Previous studies have shown that pakchoi potentially exerts inhibitory activity against cancer [8]. Phytochemical investigations have shown that flavonoids, hydroxycinnamic acids, and glucosinolate are the major bioactive ingredients of pakchoi, which possess high antioxidant activity [9,10]. e chemical compositions of pakchoi are varied with cultivating conditions, such as soils, weather, and water resources [11,12]. e growth and quality of pakchoi are significantly affected by inorganic elements, for example, selenate at low concentrations (0.5-1.0 mg/kg) could promote the growth of pakchoi and reduce cadmium contents in this vegetable [13]. e use of organic fertilizers and soil improvers, such as leonardite, might enrich the contents of macronutrients (Mg, Ca, K, and S) and micronutrients (Fe, Cu, Mn, and Zn). e classification of different pakchoi cultivars has been conducted by different methods. Wiesner reported that pakchoi origin can be classified through their glucosinolate profile using HPLC-DA-MS analysis. In the glucosinolate profile, levels of 3-butenyl glucosinolate and 2-hydroxy-3butenyl glucosinolate were considered important discriminators. By contrast, 5-methylsulfinylpentyl glucosinolate was found to be a characteristic ontogenetic variation of mature leaves of pakchoi by using PCA as the statistical model [14]. Another research also assessed the morphological and genetic diversity of pakchoi based on PCA and clustering analysis [15]. On the other hand, no previous approach categorized the geographical origins based on the characterization of elemental compositions of pakchoi.
us, this study presents a promising method to discriminate against the geographical origins of pakchoi by using ICP-MS elemental analysis coupled with statistical principal component analysis. e proposed method may be utilized for discriminating against the origins of other vegetables.  118 Sn, and 182 W in 6% ethanol/0.14 M HNO 3 was used to determine the sensitivity factors for all elements across the entire mass range for the measurement of diluted samples made in the semiquantitative mode. In case of measuring digested samples, ethanol was omitted in the calibration solution. High-purity ethanol was used for preparing matrix-matched standards. Internal standards for quantitative analysis were prepared in 6% ethanol/0.14 M HNO 3 for diluted samples and in 0.14 M HNO 3 for digested samples. Blanks for the measurement of diluted and digested samples were a 6% ethanol/0.14 M HNO 3 solution containing 50 µg/L of the internal standard and a 0.14 M HNO 3 procedure blank submitted to the microwave treatment and including the internal standard, respectively. Standards and the internal standard were prepared by appropriate dilution from 1000 mg/L standard stock solutions.

Sample Preparation and ICP-MS Measurements.
All pakchoi samples were lyophilized at −45°C for 3 days and then ground with a pulverizer to obtain find powder (<400 μm particle size) before storing in clean plastic bags. Samples (0.25 g) were weighed into a polytetrafluoroethylene (PTFE) vessel, and then, 4 mL H 2 O, 2 mL HNO 3 , and 2 mL H 2 O 2 were added. After 30 min, the sample was digested in a Mars X-press plus microwave digestion system (CEM, NC, USA). e digestion program was as follows: the sample was heated to 12°C for 15 min, held for 10 min, before heating to 160°C for 10 min, held for 10 min, and finally heated to 180°C within 10 min and held for 30 min. After cooling, the solution was diluted to 25 mL in a volumetric flask with ultrapure water. e contents were then diluted with deionized water and analyzed [16]. An Agilent 7900 ICP-MS instrument (Agilent Technologies, Tokyo, Japan) was utilized for the measurement of 42 elements in the pakchoi samples, which were 11  e analytical parameters of the ICP-MS were RF power at 1550 W, RF matching at 2.0 V, cell entrance at −40 V, cell exit of −60 V, cell energy discrimination at 5.0 V, spray chamber temperature at 2°C, argon was used as carrier gas at flowrate 1.09 L/min, and helium was used as auxiliary gas at 4.3 L/min. Data quantitation was achieved regarding matrixmatched multielement standards that had been prepared in 1% HNO 3 .
In this study, instrument detection limits were calculated using the raw intensity data from the standard and the blank (using ultrapure 2% nitric acid matrix) as per the following equation: IDL � 3SD blank x C x /(S x -S blank ), where SD blank is the standard deviation of the intensities of the multiple blank measurements, C x is the mean signal for the standard, S x is the signal for C x , and S blank is the signal for blank. Method detection limits (MDLs) were calculated as follows: MDL � IDL x constant volume/sample weight.
Continuing calibration verification standards were prepared from single element ICP standards (Merck) consisting of 20 mg/L Ca and Mg for the high standard series, and 250 µg/L Al, B, Cu, Rb, Sr, and Zn and 25 µg/L Cd, Co, Cs, Ni, Tl, and V for the low standard series. e calibration verifications were measured after every 10 samples.
Duplicates of two pakchoi samples were prepared. Possible matrix effects were checked by running an interference check sample consisting of Ca (50 mg/L), Na (100 mg/L), Mg (150 mg/L), Fe (200 mg/L), Cu (250 mg/L), and Zn (500mg/L). In addition, spike recovery tests and serial dilutions were performed on pakchoi samples. e spiked samples were prepared at concentration levels of 20 and 100 µg/L for the elements Al, Cu, and Sr and of 100 and 500 µg/L for the elements B, Mn, Rb, and Zn. A serial dilution check (1 : 10 followed by 1 : 3, thus 1 : 30 final dilution) was performed on one pakchoi.

Statistical Analysis.
e statistical analysis of the data was performed using STATISTICA 12 (Dell Software, United States). e principal component analysis (PCA) was applied to the data acquired from the ICP-MS analysis to evaluate the discriminants among 6 groups of pakchoi samples from 6 geographical locations. e outputs of multivariate statistical analysis included the scree plot to show contributions of principal components to the PCA model, the score scatterplot illustrates the separation of 6 different groups, the loading scatterplot explains the influences of the elements to the clustering, and the moving range charts illustrate the means and ranges of common distributions of variables on the cases. Table 1 shows the contents of 42 elements in pakchoi obtained from 6 areas in Hanoi, including Tien Phong, anh Da, Linh Nam, anh Xuan, Van Duc, and Van Noi. K was the most abundant element in pakchoi samples which accounted for 52-62% of the total elements. Furthermore, Ca was the second major element present in pakchoi which accounted for 30-40% of the total elements. e highest total element contents (about 21 mg/g of fresh material) were found in the samples from Van Duc and anh Xuan, whereas the samples from Linh Nam and Van Noi showed the lowest contents of elements at about 16 mg/g of fresh material. e concentration of each element in the samples varied significantly due to the differences in geographical locations.

Principal Components Analysis (PCA).
An unsupervised principal component analysis (PCA) was conducted to visualize the effects of geographical locations on elements and the discriminant of 60 pakchoi samples. Figure 2 shows the scree plot which is used to illustrate the contributions of principal components (PCs) in a PCA model. As can be seen that the first three PCs accounted for more than 60% of the total variation of samples (Table 2), we can suppose that these PCs carried major information of variables. e quality of the PCA models was represented by R 2 and Q 2 values. e R 2 X (cum) value at 0.499 and the Q 2 (cum) value at 0.617 obtained from the PCA model were established by PC1 and PC2. e result indicated that 49.9% and 61.7% of the total variation could be explained and predicted, respectively, by the first two PCs.
In the two-dimensional score scatterplot based on PC1 and PC2 (Figure 3), pakchoi samples from 6 areas were sharply separated, which highlighted the possibility to distinguish the origins of this vegetable merely by the metal elements distribution. In addition, the contribution of variables to classify was determined by the loading scatterplot ( Figure 4). e contents of 69 Ga, 157 Gd, and 153 Eu  had the highest weight on the first PC, while 56 Fe and 64 Cu showed the largest contribution to the separation on PC2.
Moving range charts illustrated the means and ranges of common distributions of variables on the cases. As can be seen, anh Da samples had significantly higher contents of 56 Fe than the samples from other areas that could be used as a marker for distinguishing ( Figure 5(b)). Meanwhile, in the samples from Linh Nam, the elemental isotopes 59 Co and 118 Sn were dramatically increased as compared to samples from other sites that could be found as two important discriminating elements for this area (Figures 5(c) and 5(j)). Whereas, a unique combination of three elements ( 28 Si, 63 Cu, and 85 Rb) was identified as the main discriminator of pakchoi from anh Xuan ( Figure 5

Journal of Analytical Methods in Chemistry
Loading scatterplot (p1 vs. p2) 11 Fe   TIEN PHONG  TIEN PHONG  TIEN PHONG  TIEN PHONG  TIEN PHONG  TIEN PHONG  TIEN PHONG  TIEN PHONG  TIEN PHONG  TIEN PHONG  THANH DA  THANH DA  THANH DA  THANH DA  THANH DA  THANH DA  THANH DA  THANH DA  THANH DA  THANH DA  LINH NAM  LINH NAM  LINH NAM  LINH NAM  LINH NAM  LINH NAM  LINH NAM  LINH NAM  LINH NAM  LINH NAM  THANH XUAN  THANH XUAN  THANH XUAN  THANH XUAN  THANH XUAN  THANH XUAN  THANH XUAN  THANH XUAN  THANH XUAN  THANH       For the Van Noi samples, the content of 93 Nb was threefold to tenth-fold higher than the levels of this element in the other samples ( Figure 5(h)). us, 93 Nb was responsible for the unique classification of samples from Van Noi. Interestingly, no elemental isotope was identified that contained at a dramatically high level in Tien Phong samples; however, it was possible to discriminate this area's samples based on the significantly low content of 69 Ga, which was also identified as a discriminator for Van Duc (Figure 5(a)). is contrast was  also seen on scores scatterplots where samples from Tien Phong and Van Duc distributed on two contrary ranges of PC1.
To the best of our knowledge, this is the first time the mineral elements are used for discrimination of pakchoi geographical origins. e chemometric-based ICP-MS approach was applied to distinguish many types of foods and plants, such as cabbage [6], chili [17], and tea [18]. PCA was used widely in those reports that could find clearly and preciously the most important elements for the classification of the analytes.

Conclusions
is study is the first report of the method to determine the geographical origin of various pakchoi samples based on ICP-MS analysis and multivariate statistical analysis (PCA). e results illustrated that eleven elemental discriminators for vegetables were identified, namely, 28 Si, 56 Fe, 59 Co, 63 Cu, 69 Ga, 75 As, 85 Rb, 93 Nb, 107 Ag, 118 Sn, and 137 Ba. By using the PCA model, 60 samples of 6 areas in Hanoi were classified clearly with one-half of the variation that could be explained. e findings of this study imply that the discrimination of pakchoi based on their geographical locations could be well achieved by the combination of elemental profiling and multivariate statistical analysis. e reported method is convenient, fast, and environmentally friendly, with potential applications in distinguishing origins of a larger amount, ranges, and types of vegetables.
Data Availability e majority of the data used to support the findings of this study are included within the article. Other data are available from the corresponding author upon request.

Conflicts of Interest
e authors declare that they have no conflicts of interest.