Chemometrics-enhanced high performance liquid chromatography-diode array detection strategy for simultaneous determination of eight co-eluted compounds in ten kinds of Chinese teas using second-order calibration method based on alternating trilinear decomposition algorithm

Journal of Chromatography A, 1364 (2014) 151–162

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Chemometrics-enhanced high performance liquid chromatography-diode array detection strategy for simultaneous determination of eight co-eluted compounds in ten kinds of Chinese teas using second-order calibration method based on alternating trilinear decomposition algorithm Xiao-Li Yin, Hai-Long Wu ∗ , Hui-Wen Gu, Xiao-Hua Zhang, Yan-Mei Sun, Yong Hu, Lu Liu, Qi-Ming Rong, Ru-Qin Yu State Key Laboratory of Chemo/Biosensing and Chemometrics, College of Chemistry and Chemical Engineering, Hunan University, Changsha 410082, China

a r t i c l e

i n f o

Article history: Received 1 July 2014 Received in revised form 20 August 2014 Accepted 21 August 2014 Available online 27 August 2014 Keywords: Chinese tea High-performance liquid chromatography-diode array detection Second-order calibration Alternating trilinear decomposition Cluster analysis

a b s t r a c t In this work, an attractive chemometrics-enhanced high performance liquid chromatography-diode array detection (HPLC-DAD) strategy was proposed for simultaneous and fast determination of eight co-eluted compounds including gallic acid, caffeine and six catechins in ten kinds of Chinese teas by using second-order calibration method based on alternating trilinear decomposition (ATLD) algorithm. This new strategy proved to be a useful tool for handling the co-eluted peaks, uncalibrated interferences and baseline drifts existing in the process of chromatographic separation, which benefited from the “second-order advantages”, making the determination of gallic acid, caffeine and six catechins in tea infusions within 8 min under a simple mobile phase condition. The average recoveries of the analytes on two selected tea samples ranged from 91.7 to 103.1% with standard deviations (SD) ranged from 1.9 to 11.9%. Figures of merit including sensitivity (SEN), selectivity (SEL), root-mean-square error of prediction (RMSEP) and limit of detection (LOD) have been calculated to validate the accuracy of the proposed method. To further confirm the reliability of the method, a multiple reaction monitoring (MRM) method based on LC–MS/MS was employed for comparison and the obtained results of both methods were consistent with each other. Furthermore, as a universal strategy, this new proposed analytical method was applied for the determination of gallic acid, caffeine and catechins in several other kinds of Chinese teas, including different levels and varieties. Finally, based on the quantitative results, principal component analysis (PCA) was used to conduct a cluster analysis for these Chinese teas. The green tea, Oolong tea and Pu-erh raw tea samples were classified successfully. All results demonstrated that the proposed method is accurate, sensitive, fast, universal and ideal for the rapid, routine analysis and discrimination of gallic acid, caffeine and catechins in Chinese tea samples. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Tea, firstly discovered in China and grown in about 30 countries, is one of the most widely consumed beverages in the world [1–3]. Numerous studies have reported the biological functions of tea, for example, anti-oxidant, anti-inflammation, anti-carcinoma, anti-obesity, anti-atherosclerotic and anti-viral properties, etc. [4–8]. These beneficial effects have been attributed mainly to the presence of polyphenols and purine alkaloids in tea [9]. There

∗ Corresponding author. Tel.: +86 731 88821818; fax: +86 731 88821818. E-mail address: [email protected] (H.-L. Wu). http://dx.doi.org/10.1016/j.chroma.2014.08.068 0021-9673/© 2014 Elsevier B.V. All rights reserved.

are already growing epidemiological and preclinical evidences showing that tea polyphenols can reduce the risks of cardiovascular diseases and a variety of other cancers (such as oral cavity, esophagus, stomach, liver, small and large intestine, and mammary gland) in humans [10,11]. The major functional components in tea are gallic acid (GA) and tea catechins, mainly (+)-catechin (C), (−)-epicatechin (EC), (−)-epigallocatechin(EGC), (−)-epicatechin gallate (ECG), (−)-epigallocatechin gallate (EGCG) and (−)-gallocatechin-gallate (GCG). In addition, caffeine (CAF), the major alkaloid, is responsible for the stimulating effect [12]. Among them, EGCG is the most abundant catechin and may represent up to 50% of the catechins by weight. The structures of gallic acid, caffeine and various catechin monomers are shown in

152

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

Fig. 1. Chemical structures of gallic acid, caffeine and six catechins.

Fig. 1. Considering these potential therapeutic effects and the great consumption of the tea worldwide, further investigation of the health-promoting function, the quality control and safety of tea have gained increasing attention. The function and quality of tea are closely related to its chemical constituents. Indeed the contents of gallic acid, caffeine and catechins greatly vary depending on the tea species, age of the leaves and environmental conditions of their production sites [13]. It is therefore essential to establish an effective and convenient analytical method for the identification and determination of polyphenols and alkaloids in various tea samples. The analytical methods used for determination of gallic acid, caffeine and catechins in teas and other biological matrices include capillary electrophoresis (CE) [1], Fourier transform near infrared spectrometry [14], liquid chromatography (LC) techniques including high-performance thin layer chromatography (HPTLC) [15], monolith column, ultrahigh-pressure liquid chromatography (UHPLC) and high-performance liquid chromatography (HPLC) coupled with UV or diode array detector (UV/DAD), electrochemical or mass spectrometry detection [16–22]. Among these analytical methods, HPLC coupled with UV/DAD is by far the most popular method for the analysis of functional components in tea due to its advantages of excellent separations, high reproducibility, sufficient low detection limit, low cost and thus can be applied for the

identification and quantification of gallic acid, purine alkaloids and catechins. However, there are still some drawbacks that should be pointed out. Firstly, almost all published HPLC separation methods require 20–90 min for per run analysis, with the average being 20–40 min, resulting in relatively low efficiency [23,24]. Secondly, because of the existence of the complex matrices and cis-trans isomers, complex chromatographic conditions and large solvents consumption are unavoidable [25]. Thirdly, owing to the different conditions of laboratories in practical applications, the reversed-phase LC methods referred to above may fail to provide satisfactory separation for the analytes from each other and/or from other compounds existing in the samples and thus quantification of gallic acid, purine alkaloids and catechins in real samples may become ambiguous and lack of universality. Finally, other problems such as baseline drifts, changes of the peaks shape as well as shifts in the elution time may also decrease the quality of the final results of the analysis. Accordingly, a simple, accurate, universal, fast and environmental-friendly approach with shorter HPLC analysis time is required to facilitate larger and more extensive studies with large sample sets. Chemometrics has been widely applied in chromatography in recent 30 years, offering robust and reliable data analytical alternatives to handle the problems derived from the instability of the chromatographic systems (baseline/background

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

correction) and the lack of sample-to-sample stability (alignment or normalization), or even the problems in quantifying overlapped peaks (co-elution) [26]. Among chemometric strategies, the multiway calibration methods based on “mathematical separation” are a shining pearl, and have been utilized in combination with second- or high-order instrument for the analysis of unresolved peaks and uncalibrated interferences in many complex samples successfully [27–31]. HPLC-DAD assisted with second-order calibration methods is a suitable alternative method to resolve co-eluted peaks even in the presence of unknown interferences (the well-known ‘second-order advantage’) and realize accurate quantification in various matrices [32]. It is possible to reduce the duration of the chromatographic separation, allowing not only processing multiple samples but also reducing the solvent consumption, which exhibits great potential to be extended as a promising alternative for more practical applications. In this work, HPLC-DAD combined with second-order calibration method based on the alternating trilinear decomposition (ATLD) algorithm was developed to identify and quantify real contents of gallic acid, caffeine and six catechins in Chinese teas. Based on the literatures, complete baseline separation is difficult to achieve under simple mobile phase conditions as highly overlapping among analytes and interferences in the complex matrices would occur. However, with the help of “mathematical separation” instead of the conventional optimization of “physical and/or chemical separation”, the ATLD method can serve as an enhanced strategy to predict accurate concentrations together with reasonable resolution of chromatographic and spectral profiles for the compounds of interest, even in the presence of unknown interferences. The baselines drifts were removed by means of regarding it as additional factor(s) as well as the analyte(s) of interest in the mathematical model. The efficiency of the proposed approach was validated by the analytical figures of merit (FOM) including SEN, SEL, LOD and the statistic parameter of RMSEP. F-test and t-test were used to compare the statistical significance of results obtained from the proposed method and MRM method based on LC–MS/MS. The repeatability and reproducibility of the proposed method were also estimated by intra- and inter-day experiments. Furthermore, as a universal method, the developed strategy was applied for the analysis of several other Chinese teas, and subsequently the quantitative results were used for the clustering of tea samples based on principal component analysis (PCA).

153

where N denotes the number of factors, which is actually the total number of detectable components of interest and the background as well as unknown interferences; xijk , the element of X, is the response intensity of sample k at elution time i and UV spectrum channel j; ain , bjn and ckn are the elements of three underlying profile matrices A (I × N), B (J × N) and C (K × N) of X, respectively; eijk represents the element of the I × J × K three-way residual array, E. 2.2. Alternating trilinear decomposition (ATLD) algorithm ATLD algorithm is used to resolve the trilinear model by utilizing alternating least-squares principle and improve the quality of trilinear decomposition solution by using Moore–Penrose generalized inverses based on truncated singular value decomposition, which is developed by Wu et al. in 1998 [33] and has the advantages of being insensitive to excessive component numbers and converging more effectively [34]. Additionally, when the ratio of signal-to-noise is proper, ATLD will give reasonable results even the collinearity in the data is high [35]. The ATLD algorithm alternately minimizes the following three objective functions (2) to (4) to update the qualitative profiles (A and B) and the relative concentrations (C) of individual components: 1 (C) =

K    X..k − A diag(c(k) )BT 2 F

(2)

k=1

2 (A) =

I    Xi.. − B diag(a(i) )CT 2 F

(3)

i=1

3 (B) =

J    X.j. − C diag(b(j) )AT 2 F

(4)

j=1

2.3. Quantitative analysis Quantification by ATLD method is performed like in traditional chromatographic analysis based on external calibration procedures. Linear relationships between peak heights (resolved relative concentrations (C) from ATLD) and concentrations (cSt ) of the analytes in standard mixture samples allow building up a calibration curve and predicting the absolute concentrations (cun ) of each individual components of interest in unknown samples (Fig. 2(d)).

2. Theory 3. Experimental

2.1. Trilinear component model for second-order calibration In an HPLC-DAD analysis, UV spectra (e.g. in the range of 190–400 nm) are obtained for each elution time scan point. In this case, each sample will generate a two-way matrix Y (I × J), where I is the number of rows corresponding to the elution time scan points and J is the number of columns corresponding to the number of selected UV spectrum channels, respectively. Fig. 2(a) showed a landscape of the HPLC-DAD data acquired from one sample (i.e., the two-way matrix Y) and such a data set obtained from multiple samples including calibration samples and prediction samples was shown in Fig. 2(b), which will form a three-way data array X (I × J × K) along the dimension of the number of analyzed samples K. If there are no changes in peak positions or shapes from sample to sample, this three-way data array will have an internally mathematical structure called trilinear (Fig. 2(c), Eq. (1)), which can be depicted as follows: xijk =

N 

ain bjn ckn + eijk ,

for i = 1, . . ., I; j = 1, . . ., J;

n=1

k = 1, . . ., K,

(1)

3.1. Reagents and solutions Analytical-reagent grade chemicals and Milli-Q water are employed in all experiments. HPLC-grade acetonitrile and methanol were obtained from Merck (Tedia, USA); phosphoric acid (85%) was purchased from Tianjin Hengxing Chemical Preparation Co. Ltd. (Tianjin, China). Standards of gallic acid (GA) (>98%), (+)-catechin (C) (98%) and (−)-epicatechin (EC) (98%) were purchased from the National Institute for Control of Biological and Pharmaceutical Products (Changsha, China). Caffeine (CAF) (99%) and (−)-gallocatechin3-gallate (GCG) (98%) were purchased from Aladdin (Shanghai, China). The standards of (−)-epigallocatechin gallate (EGCG) (99%), (−)-epigallocatechin (EGC) (98%) and (−)-epicatechin gallate (ECG) (98%) were bought from Chengdu Must Bio-technology Co., Ltd. (Chengdu, China). Stock standard solutions (102–1708 ␮g mL−1 ) of GA, EGC, CAF, C, EC, EGCG, GCG and ECG were prepared by dissolving each compound in aqueous methanol (water:methanol = 80:20, v/v). All solutions were stored at 4 ◦ C in a refrigerator and were stable over 1

154

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

Fig. 2. Scheme of trilinear decomposition and external calibration strategy to obtain quantitative information from ATLD resolved relative concentration profiles. Table 1 Concentrations (␮g mL−1 ) of eight analytes of interest in calibration samples. Calibration Samples

CAF

EGCG

GA

EGC

EC

ECG

GCG

C

1 2 3 4 5 6 7 8 9 10

5.085 15.26 30.51 40.68 50.85 61.02 71.19 80.93 91.53 101.7

25.62 68.32 102.5 136.6 172.8 8.540 51.24 85.40 119.6 153.7

1.530 3.060 4.590 0.255 2.040 3.585 5.100 1.020 2.550 4.080

36.88 73.76 4.610 46.10 82.98 13.83 55.32 92.20 27.66 64.54

29.10 58.20 23.28 52.38 17.46 46.56 8.73 40.74 2.910 34.92

36.33 15.57 51.90 31.14 7.785 46.71 25.95 2.595 41.52 20.76

7.650 5.950 4.250 2.550 0.4250 8.500 6.800 5.100 3.400 1.275

13.30 11.97 10.64 9.310 7.980 6.650 5.320 3.990 1.330 0.3325

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

month. Ten calibration samples were prepared with 2 mL volumetric flasks covering the concentration ranges as shown in Table 1. 3.2. Tea samples collection and preparation Ten commercially available Chinese tea samples were purchased from local markets in Changsha. Tea leaves were minced, and infusions were prepared by extraction of 1 g of the obtained powder with 50 mL boiling water followed by brewing in a water bath at 90 ◦ C for 10 min in dark. Tea infusions were filtered through a syringe filter (0.22 ␮m porous size) and finally appropriately diluted with aqueous methanol (water:methanol = 80:20, v/v) according to the original contents of gallic acid, caffeine and catechins. 3.3. Chromatographic instrument and conditions The separation was performed on an LC-20AT liquid chromatographic system (Shimadzu, Japan) equipped with a degasser, a pump, a manual injector provided with a 20 ␮L loop, a column oven and a diode array detector (DAD). The chromatographic separation was performed on a guard and analytical cartridge system (WondaSil C18 column, 5 ␮m average particle size, 200 mm × 4.6 mm i.d., GL Sciences Inc., Japan) with column temperature set at 30 ◦ C. The DAD acquisition wavelength was set to scan from 190 to 400 nm with a 1 nm interval and a 0.64 s cycle−1 . Samples were separated using the mobile phase consisting of water with 0.1% orthophosphoric acid (A) and acetonitrile (ACN) (B) at a flow rate of 1 mL min−1 under the gradient elution procedure. The solvent gradient programming was 16%–25% B over 0.3 min, hold at 25% B for 1.7 min, 25–45% B over 0.2 min, hold at 45% B for 1.8 min, and then back to 16% B. The injection volume was 20 ␮L. 3.4. LC–MS/MS confirmatory analysis The confirmatory LC–MS/MS analysis was performed on an HPLC–MS/MS system (Agilent 1290 Infinity LC coupled to an Agilent 6460 triple quadrupole mass spectrometer, USA) equipped with an automatic sample injector and a binary pump. The triple quadrupole system employed an electrospray ionization source (ESI). A ZORBAX Eclipse XDB-C18 (150 mm × 2.1 mm i.d., 3.5 ␮m average particle size) analytical column (Agilent Technologies, Palo Alto, USA) was selected for chromatographic separation. The mobile phase was consisted of (C) water with 0.1% formic acid and (D) acetonitrile (ACN) with 0.1% formic acid. A gradient experiment was performed from 5 to 20% D in 15 min at a flow rate of 0.3 mL min−1 with the injection volume of 5 ␮L. The optimized instrument parameters, MRM transitions and their corresponding fragment voltages and collision energies for the analysis of gallic acid, caffeine and six catechins were listed in Support Information. 3.5. Data analysis The data obtained from HPLC-DAD were loaded into MATLAB and the calculations are carried out with the ATLD code available from our laboratory. 4. Results and discussion Catechins (Fig. 1) are characterized by similar chemical structures only differing in the number and/or position of hydroxyl groups. Although some methods based on HPLC-UV/DAD offer an excellent peak resolution and can be applied for catechins identification and quantification, most of them can just separate at maximum seven catechins and with long time gradient elution [2,24]. However, when more catechins or other major chemical

155

components, such as gallic acid, caffeine, etc., need to be analyzed, these methods may no longer work well. In general, it was found that the published separation methods were somewhat or entirely irreproducible both with regard to complete resolution of the analytes and the quality of chromatography, especially when using conventional C18 columns (non-deactivated 5 mm ODS stationary phase). Besides, the reproducibility of these separations was also confounded by the absence of precise specification of stationary phases used in previous investigations [2,23]. According to the method developed by Wang et al. [36], (−)-EGC and (+)-C could not be separated in the acetonitrile/water system, while full separation could be achieved in the methanol/water system with orthophosphoric acid addition. In contrast, successful and efficient separation could be accomplished by using a C18 column and isocratic elution with the mobile phase of 0.05% orthophosphoric acid–acetonitrile (85:15, v/v) rather than the methanol-based mobile phase in the literature [37] published by Hadad. et al. In this paper, HPLC-DAD in combination with second-order calibration method based on ATLD algorithm was used with no need of complete separation and the chromatographic run time of eight analytes selected was shortened significantly (Fig. 3). Fig. 3 depicts the three-dimensional, contour and twodimensional plots of the complete landscape of UV absorption intensity as a function of elution time and wavelengths for a mixture of gallic acid, caffeine and six catechins. As shown in Fig. 3, gallic acid, caffeine and six catechins were eluted completely within 8 min. One could find that there were some regions where single analyte responded, with few interferences from other sample components, e.g., the time region of 4.30–4.80 min. However, as shown in Fig. 3c, it could be observed that C and CAF were seriously overlapped, resulting in just one peak appearing at approximately 6.0 min. It is also observed that when a tea infusion sample was analyzed (Fig. 3c), unknown interferences overlapped with GA, GCG, and there were some low intensity interferences overlapped with EC, EGCG and ECG. Moreover, the unknown interferences in other real samples are unpredictable, resulting from the differences of tea species, age of the leaves and environmental conditions of their production sites. In such cases, univariate methods could not be directly applied to quantify these overlapped analytes. Fortunately, we can turn to second-order calibration methods for help, which may provide satisfactory results for the quantification of all studied components even in the presence of overlapped peaks or unknown interferences due to the well-known ‘second-order advantage’ [34,38–46]. Moreover, second-order calibration methods are also known to provide increased sensitivity and selectivity over any zeroth-order and first-order counterparts [47,48]. 4.1. Quantification of gallic acid, caffeine and catechins in tea samples using ATLD method 4.1.1. Non-trilinear factor Non-trilinear factors, for example, changes in the shapes of the peaks and shifts in the elution time or the baseline, etc., was avoidable in the chromatographic analysis, because of the changes in the temperature of column, the flow rate, the composition of gradient elution, the matrix effects of sample, the pressure of the column and the sampler injector. When non-trilinear factors are serious, trilinear algorithms cannot fit non-trilinear data completely, because the underlying trilinear model is not fulfilled. The result will be profiles which do not correspond to real ones, leading to wrong analyte concentrations. Thus, a proper correction step is crucial to the quality of chromatograms. As shown in Fig. 3, the baseline drifts in samples can be observed. For the baseline variations, if it is serious, it may lead to a systematic error for the estimated concentrations. Therefore, in this paper, it is appropriately corrected via the procedure proposed by Zhang

156

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

Fig. 3. (a) Three-dimensional plot of a typical chromatogram of a sample containing the studied gallic acid, caffeine and six catechins of calibration samples, (b) corresponding two-dimensional contour plot, and (c) corresponding two-dimensional chromatogram of calibration and real samples ( = 210 nm).

et al. [49]. The basic idea of this strategy is to implement the drifts into the mathematical model by means of regarding the drifts as additional factor(s) as well as the target analytes. Fig. 5(A-a) showed that the baseline drifts have been eliminated successfully. Additionally, in order to achieve better results with ATLD method, other non-trilinear factors should be considered since they represent loss of trilinearity of the data sets (which is required for these models). Therefore, chromatographic traces ( = 210 nm) of 10 calibration samples and 8 tea samples (spiked samples and tea samples) have been plotted respectively in Fig. 4 for checking the non-trilinear factors (e.g. time shifts and shape changes in the chromatographic peaks) in our data sets. Just as can be seen from Fig. 4, not only the peak positions of each sample are quite similar, but

also the shapes of each peak are nearly the same, indicating that our data sets are in principle trilinear. Nevertheless, slight time shifts can be observed from the Fig. 4, since the variation of chromatographic peaks from run to run cannot be avoided. However, according to the experience in our previous work, if the time shifts are not more than 0.03 s, the ATLD algorithm can use the excessive factors to fit the small non-linear factors as interferences and give satisfactory quantitative results [30]. In this case, time shifts are little enough so that they have nearly no effect on the results obtained by ATLD model. This point will be further confirmed later in the present work. On the contrary, if the time shifts are serious, other chemometric approaches such as MCR-ALS [50], PARAFAC2 [51,52], etc. can be available.

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

157

Table 2 Correlation coefficient (R) between real normalized spectral profiles and estimated ones of gallic acid, caffeine and six catechins in two selected tea infusions. Compound

GA EGC C CAF EC EGCG GCG ECG

Fig. 4. Chromatograms ( = 210 nm) of ten calibration samples and eight tea samples (prediction samples).

4.1.2. Quantification of gallic acid, caffeine and catechins in selected tea samples In this section, ATLD method was used to establish the calibration models for the quantification of 8 analytes in the selected tea infusions. In order to improve the efficiency of the data analysis, six sub-segments were divided from the entire three-way data array according to the elution ranges of analytes. Co-elution is one of the most difficult problems in chromatographic analysis due to the complexity and diversity of matrices, as well as the similarity of chemical structure of analytes and the requirements of fast analysis. Therefore, we did not really know the underlying number of components before data treatment using multi-way calibration methods. Although the ATLD method is well known for being insensitive to excessive factors, its performances on providing accurate solutions will be certainly improved when the most appropriate factors are chosen, especially for the data sets of HPLC where slight shifts in elution time usually occur. In the present work, the ADD-ONE-UP method [53], which has a strong ability to cope with heteroscedastic noises, heavy collinearity and varying backgrounds, was used to determine the number of components. According to our experience, it deserves to employ one or two factors more than the estimated one to fit the shifts. For different teas, different numbers of factors are chosen for modeling through ATLD method. The three-way data set based on both the calibration and prediction samples were decomposed by ATLD algorithm with factor numbers selected above, and the resolved spectral profiles together with their corresponding actual ones and relative concentration modes were shown in Fig. 5. As visualized in Fig. 5, the chromatograms, spectra of the studied analytes and interferences were heavily overlapped and a large number of unknown and uncalibrated components co-eluted. Even so, our methodology can still resolve clear chromatographic and spectral profiles as well as relative concentration profiles. The resolved spectral profiles of gallic acid, caffeine and catechins matched quite well with their corresponding actual spectra, which fully exploited the “second-order advantage” irrespective to the complexity of the matrices studied. In addition, the correlation coefficient (R) between resolved profiles and actual ones for analytes were calculated via the MATLAB function ‘corrcoef’. The values of R for gallic acid, caffeine and catechins in Yueyang tippy tea and Mountain tea are shown in Table 2. The R values are close or equal to 1, indicating the accuracy and reliability of the proposed strategy. It is worth noting that the R values for C deviated from 1 a bit more seriously than other analytes. The

Correlation coefficient (R) Yueyang tippy tea

Mountain tea

0.9999 1.0000 0.9946 1.0000 1.0000 1.0000 0.9999 1.0000

1.0000 1.0000 0.9981 1.0000 1.0000 1.0000 1.0000 1.0000

possible reasons are that the peak of C is heavily overlapped with that of CAF, and CAF’s relatively higher content may also lead to the less good R value for C. The prediction results for gallic acid, caffeine and catechins based on the ATLD algorithm were summarized in Table 3, together with recoveries which were calculated from our prediction concentrations and the nominal concentrations spiked in the tea infusions. As for the results, the average recoveries were within the range of 94.8–102.1% with standard deviation (SD) less than 8.9% and 91.7–104.9% with SD less than 11.9% in Yueyang tippy tea and Mountain tea, respectively. Results showed that EGCG is the most abundant catechins in tea followed by EGC, ECG, EC, while C and GCG are the scarcest catechins among the selected tea samples, which is consistent with the results reported in literatures [36,54]. Meanwhile, the content of GA is very low and that of CAF is very rich. These results confirmed that our method is capable of accurate quantifying gallic acid, caffeine and catechins in different complex tea samples.

4.1.3. Comparison with LC–MS/MS method Originally, the HPLC method based on ISO 14502-2 [55] was selected to compare with the method proposed in this paper. However, because of the differences of the chromatographic column and tea infusions, complete separation could not be achieved under the condition of the reference method. In terms of this point, it further confirmed the necessity of second-order calibration methods in combination with chromatographic analysis, i.e., it is very difficult to simultaneously quantify multiple target analytes in complex samples with unknown interferences such as Chinese tea infusions studied in the present work only using conventional chromatographic analysis methods without any assistance of other strategies. So, according to the previous reports based on LC–MS or LC–MS/MS [18,19,22], the gradient elution conditions and the optimum MS/MS conditions described in Section 3.4 were used to analyze gallic acid, caffeine and catechins in tea infusions. (see Support Information Fig. S1). The content levels obtained for gallic acid, caffeine and catechins in tea infusions using these two methods were shown in Table 4. The results of the calibration approaches were investigated with an F-test and a t-test for gallic acid, caffeine and catechins. In 95% confidence interval, with a sample size of n1 = 3, n2 = 3, all the calculated values for F (see Table 4) were lower than the critical value (Ftable, 95% = 39.0), which meant that no significant differences existed between our proposed method and the LC–MS/MS procedure. Then, a t-test was applied [56]. The t statistic results in Table 4 were also lower than the critical value of t at the 95% confidence limit, 4.30. So we accept the null hypothesis and reach a conclusion that there were no significant differences between the proposed method and the LC–MS/MS method.

158

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

Fig. 5. Results for the analysis of calibration samples and two selected tea infusions (A: Yueyang tippy tea; B: Mountain tea). Profiles of normalized chromatograms (A-a and B-a), normalized spectra (A-b, B-b), relative concentrations (A-c, B-c) and normalized actual spectra of gallic acid, caffeine and six catechins.

4.2. Method validation 4.2.1. Repeatability and reproducibility Repeatability of analytical response was tested by performing HPLC-DAD-ATLD analysis on aliquots of three spiked infusion three times within the same day. For each infusion, relative peak intensity of the three runs was used to calculate the relative standard deviations (RSDs) for each compound. While reproducibility of analytical response was quantified by performing HPLC-DAD-ATLD analysis on freshly thawed aliquots of each infusion in triplicate on three consecutive days. For each infusion, the mean relative peak intensities of triplicate analysis

on each day were used to calculate the RSDs for each compound. As shown in Table 5, both intra- and inter-day RSDs were commonly less than 8.5% for the compounds detected in the Yueyang tippy tea and intra-day RSDs were commonly smaller than interday RSDs for all the tested samples. In particular, compounds with low responses (for example C, GCG in tea infusions) had the highest RSDs, which was consistent with our results obtained in the qualitative analysis. But on the whole, gallic acid, caffeine and most catechins exhibited relatively lower intra-day and inter-day RSDs. All these indicated that our proposed method was repeatable and reproducible.

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

159

Table 3 Results of recovery in tea infusions using ATLD method. Tea samples

Analyte

Yueyang tippy tea

GA

Original (␮g mL−1 ) 0.6833

EGC

10.65

8.047

C

CAF

32.33

EC

13.82

EGCG

34.81

2.281

GCG

27.89

ECG

Mountain tea

Spiked (␮g mL−1 )

GA

0.9102

40.86

EGC

2.128

C

CAF

39.62

EC

11.59

EGCG

65.53

1.910

GCG

ECG

13.02

Found (␮g mL−1 )

Recovery%

0.7650 1.530 2.295 6.915 13.83 20.75 0.9975 1.995 2.993 7.628 15.26 22.88 4.365 8.730 13.09 12.81 25.62 38.43 0.6375 1.275 1.913 3.893 7.785 11.68

1.178 1.806 2.779 17.29 23.59 31.65 8.946 9.995 11.25 19.97 47.04 56.81 18.44 22.47 27.11 47.72 59.73 73.33 2.892 3.531 4.311 31.73 34.48 39.73

95.7 93.3 104.3 95.9 93.6 101.2 90.1 97.7 107.1 100.2 96.4 106.9 105.8 99.1 101.5 100.8 97.3 100.2 95.7 98.0 106.2 98.4 84.7 101.4

0.7875 1.575 2.363 8.940 17.88 26.82 1.250 2.500 3.750 9.930 19.86 29.79 5.070 10.14 15.21 15.22 30.44 45.66 0.7200 1.440 2.160 4.920 9.840 14.76

1.755 2.403 3.084 49.63 59.86 66.77 3.288 4.340 5.645 50.10 59.26 70.83 16.67 20.80 25.34 83.46 94.24 115.6 2.639 3.329 4.394 18.07 23.11 26.98

107.3 94.8 92.0 98.1 106.3 96.6 92.8 88.5 93.8 105.6 98.9 104.8 99.9 90.8 90.4 117.8 94.3 109.7 101.2 98.5 114.9 102.5 102.5 94.5

4.2.2. Analytical figures of merit To evaluate the performances of the proposed method, the validation parameters including SEN, SEL [57,58], RMSEP, LOD [59,60], which depend strongly on other constituents such as

Ave (Recovery ± S.D.) %. 97.8 ± 5.8

96.9 ± 3.9

98.3 ± 8.5

101.2 ± 5.3

102.1 ± 3.4

99.4 ± 1.9

100.0 ± 5.5

94.8 ± 8.9

98.0 ± 8.1

100.3 ± 5.2

91.7 ± 2.8

103.1 ± 3.6

93.7 ± 5.4

99.4 ± 11.9

104.9 ± 8.8

99.8 ± 4.6

unknown interferences and other interesting analytes in the same retention time region, were investigated. As can be seen from Table 5, GCG showed the lowest SEL value, which may result from the fact that GCG was heavily overlapped with

Table 4 Contents of gallic acid, caffeine and six catechins (mg g−1 ) obtained by two methods in Yueyang tippy tea and Mountain tea together with the significance testing results. Compound

GA EGC C CAF EC EGCG GCG ECG

Yueyang tippy tea

Mountain tea

HPLC-DAD-ATLD

LC–MS/MS

F-testa

t-Testb

HPLC-DAD-ATLD

LC–MS/MS

F-test

t-Test

0.34 5.33 4.02 16.16 6.91 17.40 1.14 13.94

0.30 4.19 4.29 19.52 7.85 18.61 0.86 12.23

2.73 2.38 0.99 0.36 0.82 3.46 0.06 0.08

0.25 0.15 0.78 0.22 0.10 0.88 0.37 0.39

0.46 20.43 1.06 19.81 5.79 32.76 0.96 6.51

0.55 21.70 0.95 22.11 6.82 32.93 0.80 6.67

1.00 0.76 2.98 2.38 0.61 0.18 0.22 5.04

0.59 0.56 0.67 1.45 0.59 0.32 0.36 0.16

a F-test is computed by F = s12 /s22 , where s1 and s2 are the standard deviations of the two methods respectively. At the 95% confidence level, with a sample size of n = 3, the critical value for F is 39.0.   b





t-test is computed by t = x1 − x2  /



((n1 − 1)s12 + (n2 − 1)s22 )/(n1 + n2 − 2)



1/n1 + 1/n2

, where x1 and x2 are the mean of the two methods respectively, n1 and

n2 are the number of samples analyzed of the two methods respectively. At the 95% confidence level, with a sample size of n1 = 3, n2 = 3, the critical value for t is 4.30.

160

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

Table 5 Repeatability and reproducibility of peak intensity of Yueyang tippy tea and the analytical figures of merit in tea infusions obtained from ATLD method. Analyte

Yueyang tippy tea

Mountain tea

SEL

SEN (mL ng−1 )

Concentration spikea (␮g mL−1 )

RMSEP (␮g mL−1 )

LOD (␮g mL−1 )

Intra-day (RSD%)b GA EGC C CAF EC EGCG GCG ECG a b c

86.92 30.01 51.46 70.94 25.38 16.58 55.74 45.13

0.27 0.58 0.49 0.54 0.42 0.21 0.08 0.55

0.07 0.22 0.15 1.75 0.81 2.55 0.05 0.74

0.13 0.68 0.17 1.19 0.23 0.51 0.09 0.85

3.68 0.97 5.64 1.74 4.59 1.54 1.54 0.45

2.35 2.40 2.70 1.65 1.56 2.26 1.35 3.61

SEN (mL ng−1 )

SEL

LOD (␮g mL−1 )

RMSEP (␮g mL−1 )

51.73 30.63 31.55 66.14 28.87 21.35 11.23 29.67

0.22 0.63 0.64 0.67 0.49 0.28 0.05 0.37

0.19 2.42 0.09 6.18 0.61 4.22 0.04 0.50

0.15 1.03 0.27 1.09 1.23 3.85 0.23 0.60

Inter-day (RSD%)c 1.76 0.41 0.95 2.20 0.47 1.71 3.75 0.86

5.35 6.90 4.58 5.11 4.06 6.87 2.28 2.61

3.92 5.21 5.72 8.04 3.82 6.39 2.25 1.21

4.9 5.24 3.43 6.07 2.65 5.18 8.48 5.03

Concentration spiked, ␮g mL-1: the concentration level of spiking is identical to Table 3. Intra-day RSDs represent data from three analyses of each spiked infusions with the same day. Inter-day RSDs represent data from triplicate analyses of each spiked infusions performed on 3 consecutive days.

unknown interferences and had lower absorbance than other analytes.

4.3. Application of the proposed method 4.3.1. Analysis of other kinds of tea samples As a universal method, we applied the proposed strategy to determine the contents of gallic acid, caffeine and catechins in several other kinds of Chinese tea samples including different levels and varieties. The results were listed in Table 6. The quantitative results were generally in agreement with those of previous description. All kinds of tea contained high levels of EGCG, while the contents of GA, C and GCG were very low for the selected tea samples. For different levels of Biluochun, the contents of the eight marker compounds were not the same, but limited by the number of samples at hand, there were no clear correlation with their levels, more samples need to be explored. For different manufacturers, the contents of the eight analytes in two kinds of Tieguanyin tea also existed differences. The contents of catechins and caffeine of Oolong (Mountain tea, Tieguanyin tea, and Oolong tea) and Pu-erh raw tea were lower than those of green tea (Biluochun tea, Longjing tea), while C was not detected in Oolong and two kinds of Tieguanyin teas. On the contrary, the level of GA was higher in Pu-erh raw tea than that of green tea. Green tea is derived directly from inactivating by steaming or microwave and drying the fresh tea leaves, thus, the chemical compositions of green tea are very similar to those of fresh tea leaves. For Oolong tea and Puerh raw tea, which were partially fermented teas, most of catechins, are oxidized and polymerized by endogenous or microbial enzymes during the manufacturing process. Accordingly, Oolong tea and Puerh raw tea contained less catechins compared with green tea. The fermentation process also increases the liberation of GA from ECG

and EGCG, resulting in relatively high levels of GA in Oolong and Pu-erh raw tea. However, we also noted that Yueyang tippy tea, an important variety of Chinese green tea, had lower contents of catechins than that of Biluochun and Longjing teas. In a word, the contents of maker compounds of tea are associated with the geographical origins, tea species, age of the leaves and environmental conditions of their production sites [61]. Thus, in the next section, a clustering analysis was done based on our quantitative results in combination with PCA.

4.3.2. Clustering of tea varieties PCA was used to achieve reduction of dimension and to observe a primary evaluation of the between-class similarity. The contents of gallic acid, caffeine and catechins listed in Table 4 and Table 6 were used for the clustering of varieties of teas by PCA. A score plot was obtained by using some top principle components from the eight variables keeping most of the original information intact in the data set. The top five principal components represented about 99.45% of the total variances, where PC1, PC2, PC3, PC4 and PC5 contributed 42.35, 38.02, 9.04, 7.02 and 3.02% variances, respectively. As shown in Fig. 6, the green tea samples except for Yueyang tippy tea formed a group; another group was made up of Oolong tea; while Pu-erh raw tea and Yueyang tippy tea were far from both of the two groups. This once again confirmed that the chemical compositions of tea have close correlation with the geographical origins, tea species, age of the leaves and environmental conditions of their production sites. Although limited by the number of samples at hand, these results could also show that the discrimination and classification of green tea, Oolong tea and Pu-erh raw tea were possible using the contents of gallic acid, caffeine and catechins assisted with pattern recognition techniques based on PCA.

Table 6 The contents of individual gallic acid, caffeine and catechins (mg g−1 ) in leaves of eight tea cultivars. Cultivar

Mean ± S.D.a GA

Level 2 Biluo chun Sweet Biluo chun Super Biluo chun Longjing Oolong tea Tieguanyin#1b Tieguanyin#2 Pu-erh raw tea a b c

0.67 1.12 1.64 1.09 0.80 0.59 0.63 2.35

EGC ± ± ± ± ± ± ± ±

0.02 0.03 0.07 0.01 0.05 0.01 0.01 0.01

46.09 29.34 30.16 42.83 37.36 49.54 63.16 11.89

Mean content give for three determinations. Different manufacturers of Tieguanyin. Not detected.

CAF ± ± ± ± ± ± ± ±

0.07 0.534 0.21 0.44 0.28 0.05 0.13 0.05

31.34 32.78 42.40 49.97 26.25 25.54 35.76 33.93

± ± ± ± ± ± ± ±

0.57 0.16 0.36 0.67 0.42 0.43 0.12 0.05

C

EC

1.73 ± 0.01 0.91 ± 0.04 2.99 ± 0.09 0.21 ± 0.01 n.d. c n.d. n.d. 5.53 ± 0.04

7.78 5.13 7.67 12.54 5.44 11.04 13.86 17.93

EGCG ± ± ± ± ± ± ± ±

0.06 0.08 0.17 0.07 0.33 0.14 0.02 0.04

85.90 64.82 77.69 124.61 42.45 50.79 66.78 37.76

GCG ± ± ± ± ± ± ± ±

2.44 0.92 1.37 0.99 0.33 0.38 0.53 0.53

0.73 0.71 0.96 7.07 1.02 1.83 3.04 1.07

ECG ± ± ± ± ± ± ± ±

0.03 0.01 0.03 0.12 0.06 0.03 0.01 0.04

18.60 10.65 23.07 20.17 5.74 8.20 10.14 32.23

± ± ± ± ± ± ± ±

0.30 0.18 0.29 0.33 0.19 0.15 0.04 0.52

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

161

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.chroma. 2014.08.068.

References

Fig. 6. Score plot of ten tea samples in three dimensions formed by PCA.

5. Conclusions The developed chemometrics-enhanced strategy combining HPLC-DAD with second-order calibration method based on the alternating trilinear decomposition (ATLD) algorithm proved to be a simple, efficient and fast methodology for the determination of gallic acid, caffeine and six catechins in Chinese tea samples. The “second-order advantage” is a powerful characteristic of this method, which mathematically separates the interferences overlapped with the analytes and thus makes the quantitative analysis possible even in the presence of complex matrices. The baseline drifts were also removed by means of regarding the drifts as additional factor(s) as well as the analytes of interest in the mathematical model. The eight analytes were eluted within 8 min and the results of overall predictions were satisfactory and also consistent with those of LC–MS/MS. As a universal method, this developed strategy was applied for analyzing another eight tea cultivars including different varieties and levels, then PCA was used for clustering the ten different tea varieties measured in this paper successfully. Therefore, the proposed method can be a feasible alternative enhanced strategy to achieve the rapid quantification of the gallic acid, caffeine and six catechins in Chinese tea samples, and PCA or other pattern recognition methods can be used to develop a clustering model for Chinese teas based on the quantitative results, which was solvent and time savings and can reduce both of the cost per analysis and environmental pollution. At last, we should emphasize that the application in this work is only a limited example of the enormous potential of this strategy in the analysis of Chinese tea. More applications in the routine resolution and quantification of other target compounds in complex samples such as biological samples, environmental samples, food samples and so on will be possible, which can be attributed to that, in principle, the second-order calibration method can be little affected by the analytical systems as long as the data meets the trilinear component model [32,62]. Acknowledgements The authors would like to acknowledge the financial supports from the National Natural Science Foundation of China (Grant No. 21175041, J1103312 and J1210040), the National Basic Research Program of China (Grant No. 2012CB910602), and the Foundation for Innovative Research Groups of NSFC (Grant No. 21221003).

[1] H. Horie, K. Kohata, Analysis of tea components by high-performance liquid chromatography and high-performance capillary electrophoresis, J. Chromatogr. A 881 (2000) 425–438. [2] J.J. Dalluge, B.C. Nelson, Determination of tea catechins, J. Chromatogr. A 881 (2000) 411–424. [3] S. Namal Senanayake, Green tea extract: chemistry, antioxidant properties and food applications – a review, J. Funct. Foods 5 (2013) 1529–1541. [4] F. Thielecke, M. Boschmann, The potential role of green tea catechins in the prevention of the metabolic syndrome – a review, Phytochemistry 70 (2009) 11–24. [5] S. Sae-Tan, K.A. Grove, J.D. Lambert, Weight control and prevention of metabolic syndrome by green tea, Pharmacol. Res. 64 (2011) 146–154. [6] N.T. Zaveri, Green tea and its polyphenolic catechins: medicinal uses in cancer and noncancer applications, Life Sci. 78 (2006) 2073–2080. [7] M.A. Potenza, F.L. Marasciulo, M. Tarquinio, E. Tiravanti, G. Colantuono, A. Federici, J. Kim, M.J. Quon, M. Montagnani, EGCG, a green tea polyphenol, improves endothelial function and insulin sensitivity, reduces blood pressure, and protects against myocardial I/R injury in SHR, Am. J. Physiol. Endocrinol. Metab. 292 (2007) 1378–1387. [8] G.D. Norata, P. Marchesi, S. Passamonti, A. Pirillo, F. Violi, A.L. Catapano, Anti-inflammatory and anti-atherogenic effects of cathechin, caffeic acid and trans-resveratrol in apolipoprotein E deficient mice, Atherosclerosis 191 (2007) 265–271. [9] C. Cabrera, R. Artacho, R. Giménez, Beneficial effects of green tea – a review, J. Am. Coll. Nutr. 25 (2006) 79–99. [10] H. Mukhtar, N. Ahmad, Tea polyphenols: prevention of cancer and optimizing health, Am. J. Clin. Nutr. 71 (2000) 1698–1702. [11] N. Khan, H. Mukhtar, Tea polyphenols for health promotion, Life Sci. 81 (2007) 519–533. [12] H. Horie, A. Nesumi, T. Ujihara, K. Kohata, Rapid determination of caffeine in tea leaves, J. Chromatogr. A 942 (2002) 271–273. [13] Y.L. Lin, I.M. Juan, Y.L. Chen, Y.C. Liang, J.K. Lin, Composition of polyphenols in fresh tea leaves and associations of their oxygen-radical-absorbing capacity with antiproliferative actions in fibroblast cells, J. Agric. Food Chem. 44 (1996) 1387–1394. [14] Q. Chen, J. Zhao, S. Chaitep, Z. Guo, Simultaneous analysis of main catechins contents in green tea (Camellia sinensis (L.)) by Fourier transform near infrared reflectance (FT-NIR) spectroscopy, Food Chem. 113 (2009) 1272–1277. [15] I. Vovk, B. Simonovska, H. Vuorela, Separation of eight selected flavan-3-ols on cellulose thin-layer chromatographic plates, J. Chromatogr. A 1077 (2005) 188–194. [16] M. Bedner, D.L. Duewer, Dynamic calibration approach for determining catechins and gallic acid in green tea using LC–ESI/MS, Anal. Chem. 83 (2011) 6169–6176. [17] Q.H. Zhang, W.B. Wang, J. Li, Y.X. Chang, Y.F. Wang, J. Zhang, B.L. Zhang, X.M. Gao, Simultaneous determination of catechin, epicatechin and epicatechin gallate in rat plasma by LC–ESI-MS/MS for pharmacokinetic studies after oral administration of Cynomorium songaricum extract, J. Chromatogr. B 880 (2012) 168–171. [18] M. Sano, M. Tabata, M. Suzuki, M. Degawa, T. Miyase, M. Maeda-Yamamoto, Simultaneous determination of twelve tea catechins by high-performance liquid chromatography with electrochemical detection, Analyst 126 (2001) 816–820. [19] B. Hu, L. Wang, B. Zhou, X. Zhang, Y. Sun, H. Ye, L. Zhao, Q. Hu, G. Wang, X. Zeng, Efficient procedure for isolating methylated catechins from green tea and effective simultaneous analysis of ten catechins, three purine alkaloids, and gallic acid in tea by high-performance liquid chromatography with diode array detection, J. Chromatogr. A 1216 (2009) 3223–3231. [20] D. Wang, J. Lu, A. Miao, Z. Xie, D. Yang, HPLC-DAD-ESI-MS/MS analysis of polyphenols and purine alkaloids in leaves of 22 tea cultivars in China, J. Food Compos. Anal. 21 (2008) 361–369. [21] Mde.L. Mata-Bilbao, C. Andrés-Lacueva, E. Roura, O. Jáuregui, C. Torre, R.M. Lamuela-Raventós, A new LC/MS/MS rapid and sensitive method for the determination of green tea catechins and their metabolites in biological samples, J. Agric. Food Chem. 55 (2007) 8857–8863. [22] M. El-Shahawi, A. Hamza, S. Bahaffi, A. Al-Sibaai, T. Abduljabbar, Analysis of some selected catechins and caffeine in green tea by high performance liquid chromatography, Food Chem. 134 (2012) 2268–2275. [23] J.J. Dalluge, B.C. Nelson, J.B. Thomas, L.C. Sander, Selection of column and gradient elution system for the separation of catechins in green tea using highperformance liquid chromatography, J. Chromatogr. A 793 (1998) 265–274. [24] Y. Wang, Q. Li, Q. Wang, Y. Li, J. Ling, L. Liu, X. Chen, K. Bi, Simultaneous determination of seven bioactive components in Oolong tea camellia sinensis: quality control by chemical composition and HPLC fingerprints, J. Agric. Food Chem. 60 (2011) 256–260.

162

X.-L. Yin et al. / J. Chromatogr. A 1364 (2014) 151–162

[25] A.P. Neilson, R.J. Green, K.V. Wood, M.G. Ferruzzi, High-throughput analysis of catechins and theaflavins by high performance liquid chromatography with diode array detection, J. Chromatogr. A 1132 (2006) 132–140. [26] J.M. Amigo, T. Skov, R. Bro, Chromathography: solving chromatographic issues with mathematical models and intuitive graphics, Chem. Rev. 110 (2010) 4582–4605. [27] J. Zhao, H.L. Wu, J.F. Niu, Y.J. Yu, L.L. Yu, C. Kang, Q. Li, X.H. Zhang, R.Q. Yu, Chemometric resolution of coeluting peaks of eleven antihypertensives from multiple classes in high performance liquid chromatography: a comprehensive research in human serum, health product and Chinese patent medicine samples, J. Chromatogr. B 902 (2012) 96–107. [28] X.H. Zhang, H.L. Wu, J.Y. Wang, D.Z. Tu, C. Kang, J. Zhao, Y. Chen, X.X. Miu, R.Q. Yu, Fast HPLC-DAD quantification of nine polyphenols in honey by using second-order calibration method based on trilinear decomposition algorithm, Food Chem. 138 (2013) 62–69. [29] J.Y. Wang, H.L. Wu, Y. Chen, Y.M. Sun, Y.J. Yu, X.H. Zhang, R.Q. Yu, Fast analysis of synthetic antioxidants in edible vegetable oil using trilinear component modeling of liquid chromatography–diode array detection data, J. Chromatogr. A 1264 (2012) 63–71. [30] Z.C. Liu, S.W. Cai, X.G. Shao, High-throughput approach for analysis of multicomponent gas chromatographic–mass spectrometric signals, J. Chromatogr. A 1216 (2009) 1469–1475. [31] V. Boeris, J.A. Arancibia, A.C. Olivieri, Determination of five pesticides in juice, fruit and vegetable samples by means of liquid chromatography combined with multivariate curve resolution, Anal. Chim. Acta 814 (2014) 23–30. [32] X.D. Qing, H.L. Wu, C.C. Nie, X.F. Yan, Y.N. Li, J.Y. Wang, R.Q. Yu, Simultaneous determination of plant growth regulators in environmental samples using chemometrics-assisted excitation-emission matrix fluorescence: experimental study on the prediction quality of second-order calibration method, Talanta 103 (2013) 86–94. [33] H.L. Wu, M. Shibukawa, K. Oguma, An alternating trilinear decomposition algorithm with application to calibration of HPLC-DAD for simultaneous determination of overlapped chlorinated aromatic hydrocarbons, J. Chemom. 12 (1998) 1–26. [34] H.L. Wu, J.F. Nie, Y.J. Yu, R.Q. Yu, Multi-way chemometric methodologies and applications: a central summary of our research work, Anal. Chim. Acta 650 (2009) 131–142. [35] Y.J. Yu, H.L. Wu, J.F. Nie, S.R. Zhang, S.F. Li, Y.N. Li, S.H. Zhu, R.Q. Yu, A comparison of several trilinear second-order calibration algorithms, Chemom. Intell. Lab. Syst. 106 (2011) 93–107. [36] H. Wang, K. Helliwell, X. You, Isocratic elution system for the determination of catechins, caffeine and gallic acid in green tea using HPLC, Food Chem. 68 (2000) 115–121. [37] G.M. Hadad, R.A. Abdel Salam, R.M. Soliman, M.K. Mesbah, Rapid and simultaneous determination of antioxidant markers and caffeine in commercial teas and dietary supplements by HPLC-DAD, Talanta 101 (2012) 38–44. [38] H.L. Wu, Y. Li, R.Q. Yu, Recent developments of chemical multiway calibration methodologies with second-order or higher-order advantages, J. Chemom. 28 (2014) 476–489. [39] V. Gómez, M.P. Callao, Analytical applications of second-order calibration methods, Anal. Chim. Acta 627 (2008) 169–183. ˜ [40] J.A. Arancibia, P.C. Damiani, G.M. Escandar, G.A. Ibanez, A.C. Olivieri, A review on second- and third-order multivariate calibration applied to chromatographic data, J. Chromatogr. B 910 (2012) 22–30. ˜ de la Pena, ˜ A.C. Olivieri, Second- and [41] G.M. Escandar, H.C. Goicoechea, A. Munoz higher-order data generation and calibration: a tutorial, Anal. Chim. Acta 806 (2014) 8–26. [42] H.W. Gu, H.L. Wu, Y.J. Liu, X.L. Yin, Z.Y. Su, S.S. Li, R.Q. Yu, Simultaneous determination of metoprolol and ␣-hydroxymetoprolol in human plasma using excitation–emission matrix fluorescence coupled with second-order calibration methods, Bioanalysis 4 (2012) 2781–2793.

[43] F.Y. Tan, C. Tan, A.P. Zhao, M.L. Li, Simultaneous determination of free amino acid content in tea infusions by using high-performance liquid chromatography with fluorescence detection coupled with alternating penalty trilinear decomposition algorithm, J. Agric. Food Chem. 59 (2011) 10839– 10847. [44] M. Vosough, H.M. Esfahani, Fast HPLC-DAD quantification procedure for selected sulfonamids, metronidazole and chloramphenicol in wastewaters using second-order calibration based on MCR-ALS, Talanta 113 (2013) 68–75. [45] Y.M. Sun, H.L. Wu, J.Y. Wang, Z. Liu, M. Zhai, R.Q. Yu, Simultaneous determination of eight flavonoids in propolis using chemometrics-assisted high performance liquid chromatography-diode array detection, J. Chromatogr. B 962 (2014) 59–67. [46] X.L. Yin, H.L. Wu, X.H. Zhang, H.W. Gu, R.Q. Yu, Simultaneous determination of irinotecan and its metabolite 7-ethyl-10-hydroxycamptothecin in biological fluids using excitation–emission matrix fluorescence coupled with secondorder calibration method, Acta Chim. Sinica 71 (2013) 560–566. [47] A.C. Olivieri, Analytical advantages of multivariate data processing. One, two, three, infinity? Anal. Chem. 80 (2008) 5713–5720. [48] J.A. Arancibia, A.C. Olivieri, G.M. Escandar, First- and second-order multivariate calibration applied to biological samples: determination of antiinflammatories in serum and urine, Anal. Bioanal. Chem. 374 (2002) 451–459. [49] Y. Zhang, H.L. Wu, A. Xia, L.H. Hu, H.F. Zou, R.Q. Yu, Trilinear decomposition method applied to removal of three-dimensional background drift in comprehensive two-dimensional separation data, J. Chromatogr. A 1167 (2007) 178–183. [50] R. Tauler, Multivariate curve resolution applied to second-order data, Chemom. Intell. Lab. Syst. 30 (1995) 133–146. [51] H.A. Kiers, J.M. Ten Berge, R. Bro, PARAFAC2 – Part I. A direct fitting algorithm for the PARAFAC2 model, J. Chemom. 13 (1999) 275–294. [52] R. Bro, C.A. Andersson, H.A. Kiers, PARAFAC2 – Part II. Modeling chromatographic data with retention time shifts, J. Chemom. 13 (1999) 295–309. [53] Z.P. Chen, Z. Liu, Y.Z. Cao, R.Q. Yu, Efficient way to estimate the optimum number of factors for trilinear decomposition, Anal. Chim. Acta 444 (2001) 295–307. [54] R. Wang, W. Zhou, X. Jiang, Reaction kinetics of degradation and epimerization of epigallocatechin gallate (EGCG) in aqueous system over a wide temperature range, J. Agric. Food Chem. 56 (2008) 2694–2701. [55] BS ISO 14502-2:2005. Determination of substances characteristic of green and black tea. (Part 2: Content of catechins in green tea method using highperformance liquid chromatography), 2006. [56] P. Gemperline, Practical guide to chemometrics, second ed., CRC Press, New York, 2006. [57] A.C. Olivieri, N.K.M. Faber, A closed-form expression for computing the sensitivity in second-order bilinear calibration, J. Chemom. 19 (2005) 583–592. [58] A.C. Olivieri, K. Faber, New developments for the sensitivity estimation in fourway calibration with the quadrilinear parallel factor model, Anal. Chem. 84 (2011) 186–193. ˜ ˜ A. Espinosa Mansilla, D. González Gómez, A.C. Olivieri, [59] A. Munoz de la Pena, H.C. Goicoechea, Interference-free analysis using three-way fluorescence data and the parallel factor model. Determination of fluoroquinolone antibiotics in human serum, Anal. Chem. 75 (2003) 2640–2646. [60] A.C. Olivieri, N.K.M. Faber, Standard error of prediction in parallel factor analysis of three-way data, Chemom. Intell. Lab. Syst. 70 (2004) 75–82. [61] N.S. Ye, L.Q. Zhang, X.X. Gu, Classification of maojian teas from different geographical origins by micellar electrokinetic chromatography and pattern recognition techniques, Anal. Sci. 2 (2011) 765–769. [62] Y.N. Li, H.L. Wu, J.F. Nie, S.F. Li, Y.J. Yu, S.R. Zhang, R.Q. Yu, Interference-free determination of abscisic acid and gibberellin in plant samples using excitationemission matrix fluorescence based on oxidation derivatization coupled with second-order calibration methods, Anal. Methods 1 (2009) 115–122.