Zongde Jiang, Zisheng Han, Minghun Wen, Chi-Tang Hob,, You Wu, Yijun Wang,Na Xu, Zhongwen Xie, Jinsong Zhang, Liang Zhang,*, Xiaohun Wan,*

a State Key Laboratory of Tea Plant Biology and Utilization, Anhui Agricultural University, Hefei 230036, China

b International Joint Laboratory on Tea Chemistry and Health Effects of Ministry of Education, Anhui Agricultural University, Hefei 230036, China

c Department of Food Science, Rutgers University, New Brunswick, NJ 08901, USA

Keywords:

Taste

Roasting

Metabolomics

Correlation coefficient

Quinic acid

A B S T R A C T

Roasting is a common manufacture technology for processing various teas.It is not only used in decreasing the water content offinished tea, but also improving the flavor of teas.In the present study, the roasted and non-roasted teas were compared by liquid-chromatography mass spectrometry and sensory evaluation.The roasted tea tasted less bitter and astringent.The content of main galloylated and simple catechins, caffeine and theobromine in roasted were significantly lower than non-roasted teas.Targeted taste-compounds metabolomics revealed that (-)-epigallocatechin gallate, kaempferol-glucose-rhamnose-glucose and (-)-epicatechin gallate were main contributors tightly correlated to astringent intensity.Flavonol glycosides including kaempferol-glucose,quercetin-glucose, kaempferol-glucose-rhamnose-glucose, and quercetin-glucose-rhamnose-glucose in roasted teas were also significantly less than non-roasted teas.To study the chemical changes during roasting, tea with a strong astringency was roasted under 80, 100, 120, 140, and 160 °C.With the increase of roasting temperature,the bitter and astringent intensity of tea was gradually decreased, but the main astringent compounds including(-)-epigallocatechin, (-)-epigallocatechin gallate and kaempferol/quercetin glycosides were irregularly varied with temperature.The Pearson correlation coefficient analysis suggested procyanidin B2, coumaroylquinic acids and gallotannins were tightly correlated to the astringent and bitter perceptions, while N-ethyl-2-pyrrolidonesubstituted flavan-3-ols were negatively correlated.

1.Introduction

Like coffee, roasting is also one of the most important processing steps during production of tea [1].The applied temperature of roasting on tea is usually about 80-150 °C with an adjustable range of roasting time.There are some typical types of roasted tea in China,such as Wuyi Rock tea, large-leaf yellow tea and Luan guapian [2].After roasting, the flavor of tea shows remarkable changes, including the aroma type, taste perceptions and also the appearance of dry tea and tea infusion [3,4].The main compounds of tea raw materials are flavan-3-ols, purine alkaloids, theanine and flavonol glycosides,but after roasting some main compounds were quantitatively or qualitatively changed [5,6].For example, the (2R,3R)-flavan-3-ols were isomerized into (2S,3R)-flavan-3-ols, like the transformation of (-)-epigallocatechin gallate (EGCG) into (-)-gallocatechin gallate(GCG) [7,8].Furthermore, the volatile compounds contributed to the special caramel aroma of roasted tea during roasting [9].After roasting, the content of theanine was highly decreased, and it may involve in the formation of pyrazines volatiles.The Strecker degradation products of theanine also conjugated with catechins to formN-ethyl-2-pyrrolidone-substituted flavan-3-ols [10-12].It was suggested that roasting possessed a huge contribution to the flavor and biological activities of roasted tea.

Recently, the flavor chemistry of tea has attracted much attention besides its health benefits.The main taste compounds of tea were non-volatile compounds, like tea polyphenols, caffeine and some amino acids (L-theanine) [13].Most of known non-volatile metabolites are with the content of 1-200 mg/g in dry tea, but some non-volatile metabolites are low-threshold astringent compounds.Therefore, during roasting, the minor and trace metabolites’ changes were hardly detected by regular quantitative analysis.The untargeted metabolomics has a main advantage of detecting the minor and trace metabolites [14].

In addition, the variation of metabolites also triggers the changes of tea flavor, especially for the taste of tea, like bitterness,astringency and umami [15,16].The loss of theanine after roasting is adverse for the umami taste of green tea, but the decreasing of galloylated catechins could lower the astringency of tea infusion [17].Therefore, the aim of the present study is to compare the chemical profiles of roasted teas based on a combined strategy using liquid chromatography tandem mass spectrometry (LC-MS) analysis and an untargeted and pseudotargeted metabolomics approach [18].The global characterization of components, such as flavan-3-ols, purine alkaloids, phenolic acid, hydrolysable tannins, condensed tannins,flavonol glycosides was carried out using ultra-high-performance liquid chromatography (UHPLC) coupled with a high-resolution mass spectrometer method.Then, the combination of the untargeted and pseudotargeted metabolomics approach was used to rapidly characterize a broad array of compounds responding to chemical variations of tea during roasting.

2.Material and methods

2.1 Samples and chemicals

A total of 8 samples of roasted and non-roasted teas were obtained from different province, including 4 roasted teas: Large-leaf yellow tea (RT1), Feng huang dan cong (RT2), Strong flavor Tie guanyin(RT3), Da hongpao (RT4); and 4 non-roasted teas: Huo shan huang ya (NRT1), Jun shan yin zhen (NRT2), Lu shan yun wu (NRT3),and Gu zhang mao jian (NRT4).All samples were stored at 4 °C before analysis.Standards for gallic acid (GA, > 98%), theobromine(THB, > 98%), caffeine (CAF, > 98%), EGCG > 98%, GCG >98%, (-)-epicatechin gallate (ECG, > 98%), (-)-epicatechin (EC, >98%), (+)-catechin (C, > 98%), (-)-epigallocatechin (EGC, > 98%),(-)-gallocatechin (GC, > 98%), and rutin (> 98%), were purchased from Yuanye Biotechnology Company (Shanghai, China).HPLC and LC-MS-grade acetonitrile, methanol and water were purchased from Thermo Fisher Scientific Co.(Fair Lawn, NJ, USA).The other reagents were of analytical grade.

2.2 Sensory evaluation

On the basis of the traditional Chinese sensory evaluation protocol for tea (GB/T 23776-2018 [19]), the bitterness and astringency intensity of roasted tea and non-roasted tea were evaluated by seven trained panelists.Before evaluation, all panelists were trained 6 times in 2 weeks.To quantitatively evaluate the bitterness and astringency intensity more precisely, EGCG aqueous solution (in distilled water) at various concentrations (0, 2, 4, 6, 8 and 10 mmol/L) were used and the scores were defined as 0, 2, 4, 6,8 and 10, respectively [20,21].

2.3 Determination of main compounds

One hundred milligrams of each tea sample was extracted twice with 4 mL of water under water bath at 90 °C for 15 min.The extracts were centrifuged at 4 293 ×gfor 5 min.All supernatant was pooled and diluted to 10 mL.Then, before the HPLC and LC-MS analysis,the content was filtered using a 0.22 μm membrane.Determination of main compounds was conducted by the Agilent 1260 Infinity HPLC system (Agilent Technologies, Palo Alto, CA, USA), The gradient elution, instrument parameters and metabolomics analysis methods were used according to the previous published paper [22].

2.4 Untargeted metabolomics analysis

Untargeted metabolomics analysis was conducted according to previously published research with minor modifications [5].Chromatographic separation was achieved through the use of Acquity UPLC shield RP-18 column (50 mm × 2.1 mm, 1.7 μm) equipped with an Acquity UPLC C18guard column (Waters, Milford, MA,USA) at a flow rate of 0.3 mL/min.The column temperature and detection wavelength were set at 40 °C and 278 nm, respectively.The injection volume was 3 μL.The mobile phase consisted of 0.1%formic acid/water (V/V, A) and acetonitrile (B).The gradient elution was 0-5 min, 95%-85% A, 5%-15% B; 5-8 min, 85%-70% A,15%-30% B; 8-13 min, 70% A, 30% B; 13-23 min, 70%-12% A,30%-88% B; 23-28 min, 12%-5% A, 88%-95% B; 28-30 min, 5% A,95% B; 30-33 min, 5%-95% A, 95%-5% B; and 33-35 min, 95% A,5% B.The mass spectrometry parameters were set as follows:sheath gas flow, 11 L/min; gas flow, 8 L/min; and nebulizer, 35 psi;gas temperature, 320 °C; sheath gas temperature, 350 °C; capillary voltage, 3 500 V.The mass scan range wasm/z100-1 300 in negative ionization mode.The gradient elution, instrument parameters and metabolomics analysis was the same as previous published [23].

2.5 Targeted metabolomics analysis for taste compounds

Method of targeted metabolomics analysis of the chromatographic instrument parameters was the same as untargeted metabolomics analysis.The secondary debris information of MS/MS was set as follows: sheath gas flow, 11 L/min; gas flow, 8 L/min; and nebulizer,35 psi; gas temperature, 320 °C; sheath gas temperature, 350 °C;capillary voltage, 3 500 V; fragmentor 100 V; skimmer 65 V.The MS scan range wasm/z100-1 300 and MS/MS scan range wasm/z50-1 300 in negative ionization mode.After processing the raw data of untargeted metabolomics, a series of important taste compounds were obtained.Liquid chromatography-quadrupole-time of flight mass spectrometry (LC-Q-TOF-MS/MS) mode was employed to detect the targeted taste compounds [23].The collision energies (CE) were set up for the parent ion atm/z117.01 (0.74 min, CE 20 mV), 169.01 (1.09 min,CE 20 mV), 173.04 (0.62 min, CE 20 mV), 289.07 (3.22 min,CE 30 mV), 289.07 (4.12 min, CE 30 mV), 305.06 (1.95 min,CE 30 mV), 305.06 (2.82 min, CE 30 mV), 343.06 (0.65 min, CE 20 mV),441.08 (6.99 min, CE 30 mV), 441.08 (7.46 min, CE 30 mV),447.09 (7.72 min, CE 40 mV), 447.09 (7.89 min, CE 40 mV),457.07 (5.58 min, CE 20 mV), 457.07 (6.40 min, CE 20 mV), 463.08(7.32 min, CE 30 mV), 479.05 (6.39 min, CE 30 mV), 479.07(6.54 min, CE 30 mV), 593.15 (7.68 min, CE 40 mV), 609.14(7.18 min, CE 40 mV), 625.13 (6.49 min, CE 40 mV), 755.20(7.13 min, CE 40 mV), 755.20 (7.40 min, CE 40 mV), 771.19(6.78 min, CE 50 mV), and 771.19 (6.99 min, CE 50 mV).

2.6 Sensory evaluation and astringency-targeted metabolomics of tea during model-roasting

The NRT4 tea was roasted at 80, 100, 120, 140, and 160 °C for 60 min, respectively.Then all roasted tea samples were evaluated by trained panelists.The sensory evaluation method was the same as above.

2.7 Correlation analysis of components and astringency/bitterness

After normalizing the original data, GraphPad Prism 7.0 was used for correlation analysis,P-values below 5% were considered significant.

2.8 Statistical Analysis

All samples were tested in original triplicates, the results were expressed as mean ± standard deviation, and the statistics analysis was achieved using one-way ANOVA andt-test.Values in the same table and figure that were labeled with different letters represent a significant difference (P＜ 0.05).Untargeted metabolomics MS raw data was processed through MS-DIAL (version 4.20) and SIMCA-P(version 14.1, Umetrics, Umea, Sweden).The correlation analysis between taste compounds and astringency was achieved by using GraphPad Prism 7.0,P-values below 5% were considered significant.

3.Results and discussion

3.1 The taste of roasted and non-roasted teas

The roasted tea samples are typical “gaohuo” (high-temperature roasting) variety, the processing of which usually has a critical roasting step for shaping the final tea flavor characteristics.As shown in Fig.1, the non-roasted teas have significant higher bitter and astringent scores than roasted teas.The raw materials of roasted teas are mainly large, mature leaves with or without connected stems.Therefore, the content of main astringent compounds in the raw materials of roasted teas should be higher than those of non-roasted teas.After roasting, tea showed less astringency and bitterness, which may be explained by the heat-derived transformation of galloylated catechins or other compounds.

Fig.1 The (A) bitter and (B) astringent scores of roasted and non-roasted teas.Data are expressed as mean (n = 3); the levels of significant difference in*P ＜ 0.05, **P ＜ 0.01 and ***P ＜ 0.001 between RT and NRT.

3.2 The content of main compounds

As shown in Fig.2, no matter in roasted or non-roasted teas,the predominant compound was EGCG, which was about 45 mg/g to 75 mg/g in non-roasted teas, but 18 mg/g to 55 mg/g for roasted tea samples.Furthermore, EGC, ECG, EC, GCG, GC, GA were determined in the order from high to low concentration.Among these compounds, galloylated catechins including EGCG, ECG and GCG mainly tasted astringent because they contained the galloyl moiety in molecule.The content of EGCG and ECG in non-roasted tea samples were significantly higher than those in roasted teas (P< 0.001), but GCG’s level in roasted teas were significantly higher compared with non-roasted teas.

Fig.2 The content of main compounds of tea with or without roasting.Data are expressed as mean (n = 3, mg/g).A, GA; B, GC; C, EGC; D, C; E, THB; F, EC; G,EGCG; H, GCG; I, ECG; J, CAF.

Fig.2 (Continued)

We studied the effects of roasting on the configuration of flavan-3-ols.The epimerization of catechins were promoted by hightemperature roasting.In the present study, the roasted teas contained significant higher content of GCG, because it was mainly isomerized by heating during roasting from EGCG.The astringent threshold of EGCG is about 190 μmol/L (0.087 mg/mL).In the roasted and non-roasted teas, the concentrations of EGCG were about 0.18 and 0.75 mg/mL, both of which were significantly higher than the astringent threshold of EGCG [24].The highest content of GCG was about 0.08 mg/mL, but the threshold of GCG was 0.18 mg/mL.This result suggested that GCG may contribute less to the astringency of tea infusion.In addition, the content of EC, C, GC and GCG were lower than their astringent thresholds in all tea infusions, the content of GA was also much lower than its umami threshold.The main tasty contributors should be EGCG, EGC, ECG and caffeine.

3.3 Untargeted LC-MS based metabolomics

In total 5 545 ions were detected for all samples, and theirm/zvalues, retention time (min), mass intensity were aligned for all data(Table 1).As shown in Fig.3, the hierarchical cluster analysis (HCA)gave a clear classification for two types of teas, which suggested unsupervised multivariate analysis could effectively distinguish roasted and non-roasted teas.Furthermore, the principal component analysis (PCA), partial least square discriminant analysis (PLS-DA)and orthogonal PLS-DA (OPLS-DA) also obtained the same classification for all tea samples.To search for the critical metabolites responsible for the classification of teas, or the varied compounds which were susceptible to roasting (heating), the variable importance for the projection (VIP) compounds andS-plot were comprehensively evaluated.In Table 1, some critical compounds which were significantly different between roasted and non-roasted teas were identified by authentic chemicals or deduced by mass fragment ions with reference to tea metabolite database (TMDB) (http://pcsb.ahau.edu.cn:8080/TCDB/f).Twenty-one compounds in Table 1 were identified with chemical standards at the same retention time and quasi-molecular ion of negative mode (M-H).

Fig.3 The metabolomics and multivariate analysis for roasted and non-roasted teas by LC-MS.A, HCA; B, PCA; C, PLS scattering plot; D, OPLS-DA; E, S-plot by PLS-analysis on two types of teas (NRT vs RT); F, permutation test results.

Table 1The critical compounds responsible for the classification between RT and NRT by LC-MSn.

Table 1 (Continued)

Other critical compounds were tentatively identified mainly by high-resolution mass spectrum and mass fragment ions (MS2).Because there have been over 1 000 organic compounds reported in Reaxys database regardingCamelliaspecies and TMDB, so most of these metabolites could be deduced by regular mass fragment ions from homogeneous structural skeleton.For example, the typical MS2ion atm/z191 is the [M-H]of quinic acid [25], the MS2ion atm/zof 169 is [M-H]of GA or galloyl moiety dissociated from galloylated catechins or hydrolysable tannins molecule [26].In Table 1, there are 3 gallotannins including di-galloyl-HHDP-glucose, trigalloyl-glucose and digalloyl-glucose tentatively deduced by regular neutral loss of 170 (GA) and 152 (GA-H2O).Furthermore, some flavonol glycosides are tentatively identified by aglycones of quercetin and kaempferol atm/z301 and 285 respectively, and the neutral loss of 162 and 146 of pentose and hexose [27].The 3/4/5-p-coumaroylquinic acids were identified by the elution order in reversed phase column chromatography,and the main MS2ionsm/zat 191, 173, 163 and 119 [28].

3.4 Astringency-targeted LC-MS based metabolomics

After untargeted metabolomics analysis, all the known taste compounds of tea were selected as targeted components.All taste compounds were identified by chemical standards or LC-MS/MS fragment ions.The parent ionm/zof each compound was fragmented by collision-induced dissociation, and the product of each parent ion was integrated by deconvolution.All peak areas of taste compounds were imported into SMICA-P 14.0 software for multivariate analysis.As shown in Fig.4, the roasted and non-roasted teas could be clearly distinguished, only one tea sample (Fenghuang dancong) was reclassified into roasted tea.The contribution of each compound to re-classification were also listed in Fig.4.The first 4 important variables for the classification of non-roasted and roasted teas were EGCG,kaempferol-glucose-rhamnose-glucose (K-G-R-G), ECG and GCG.All these compounds belong to the low-threshold astringent components in tea.We also did a semi-quantitative analysis for all flavonol glycosides between roasted and non-roasted teas [23], as shown in Fig.5.

Fig.4 The metabolomics analysis of taste-related compounds in different teas (roasted and non-roasted teas).A, PLS scattering plot; B, main taste-related compounds.Q-G (quercetin-glucose); K-R-G (kaempferol-rhamnose-glucose); Q-R-G (quercetin-rhamnose-glucose); M-R-G (myricetin-rhamnose-glucose); L-TN (L-theanine);S-A (succinic acid); K-G (kaempferol-glucose); M-G (myricetin-glucose); TG (theogallin).# indicated isomers.

Fig.5 The content of flavonol glycosides (μg/mL) in roasted and non-roasted teas.n = 3, the levels of significant difference in **P ＜ 0.0.1 and***P ＜ 0.001 between RT and NRT.

Among these compounds, the content of total flavonol glycosides(TFGs) was different between two types of teas, and the content of K-G-R-G in roasted group was significantly lower than that of non-roasted teas.Other flavonols including quercetin-glucoserhamnose-glucose (Q-G-R-G), K-G-R-G#, were also low in roasted teas.Because the roasted teas were collected from different production areas and usually processed from differentCamelliacultivars, therefore the comparison on various teas under different degree of roasting preferred to a thermal model reaction with various temperature for the same tea.

3.5 The effects of roasting on the taste compounds of tea by thermal model reaction

As shown in Fig.6, the sensory evaluation on roasted teas through different temperatures were conducted by trained panelists.With the increase of roasting temperature (80-160 °C), the bitter and astringent scores of tea samples were gradually decreased.The temperature at 100 °C is a critical point for effectively decreasing the astringent intensity of tea.In addition, during sensory evaluation, we found that the burnt taste of the G6 sample was significantly increased, which may be caused by the extreme high temperature.

Fig.6 The (A) bitter and (B) astringent scores of tea under roasting with different temperature.Data are expressed as mean (n = 3); values in the bargraph above different letters are different significantly (P ＜ 0.05).G1–G6 were used to roasted model tea, G1 (non-roasted), G2 (80 °C, 60 min), G3 (100 °C,60 min), G4 (120 °C, 60 min), G5 (140 °C, 60 min), G6 (160 °C, 60 min).

Except for the taste evaluation of roasted teas, the content of main polyphenols and purine alkaloids was also quantitatively measured.The concentration of compounds in tea infusion exceeding their thresholds were labeled with dark column.GA was gradually accumulated and sharply increased after 160 °C roasting.Although the content of EGC, EGCG decreased after high-temperature roasting,there was an inverse result between 80 °C and 100 °C roasting.According to our previous report, at high-temperature roasting, the epimerization of (2R,3R)-epicatechin promoted the increasing of GCG, GC and (+)-catechin [24].However, the concentrations of these compounds were lower than respective taste threshold (Fig.7).It was suggested that these compounds whose levels were below taste thresholds would contribute less or participate in the taste formation of tea infusion by a comprehensive effect.

Fig.7 The changes of main compounds of tea under roasting with different temperature.Data are expressed as mean, n = 3.The dark column with GA, GC, EGC, C,THB, EC, EGCG, GCG, ECG, and CAF indicate that it exceeds the taste threshold, and the gray column indicate that it does not exceed the taste threshold.A, GA;B, GC; C, EGC; D, C; E, THB; F, EC; G, EGCG; H, GCG; I, ECG; J, CAF; K, total catechins; L, ester type catechins/total catechins.

Fig.7 (Continued)

Flavonol glycosides have been reported as very low-threshold astringent compounds of black tea [24].In the present study, almost all flavonol glycosides were semi-quantitatively determined by LC-MS/MS.The changes of flavonol glycosides were not tightly correlated to the roasting temperature except at 160 °C roasting.As shown in Fig.8, under the high-temperature roasting, the contents of kaempferol-glucose (K-G), quercetin-glucose (Q-G), Q-G-R-G and K-G-R-G significantly decreased.The kaempferol and quercetin di-glycosides were firstly increased but subsequently decreased along the increasing of roasting temperature.It suggested the degradation or hydrolysis of flavonoltri-glycosides may produce some flavonol di-glycosides.

Fig.8 The changes of astringent flavonol glycosides of tea under roasting with different temperatures.Data are expressed as mean, n = 3.The dark column indicated that it exceeds the taste threshold and the gray column indicates that it does not exceed the taste threshold.A, K-G; B, K-G#; C, Q-G; D, M-G; E, M-G#; F,K-R-G; G, Q-R-G; H, M-R-G; I, K-G-R-G; J, K-G-R-G#; K, Q-G-R-G; L, Q-G-R-G#.

The critical compounds which were susceptible to the roasting degree have been identified by multivariate analysis combining LCMS based metabolomics in Table 2.Some compounds were identified by authentic standards, and others were identified by mass fragment ions and tea chemistry database.

Table 2The critical compounds varied significantly during the roasting of tea.

The critical compounds (markers) related to roasting temperature were also analyzed regarding their contribution to bitter and astringent score of tea infusion.As shown in Fig.9, all taste-related compounds were classified into 8 groups (cluster I–VIII).Most of these tasterelated compounds were decreased under high-termperature roasting,but GA, GC, C, GCG,N-ethyl-2-pyrrolidione-substituted flavan-3-ols, dihydromyriciten, 4-coumaroylquinic acid, phloroglucinol and dihydroxy-methoxy-benzonic acid were increased after hightermperature roasting.

Fig.9 The heatmap and Pearson correlation coefficients between different compounds and astringent/bitter intensity.Cluster I, flavanols; Cluster II, flavonol glycosides; Cluster III, N-ethyl-2-pyrrolidione- substituted flavan-3-ols; Cluster VI, condensed tannins; Cluster V, quinic acid class; Cluster VI, small molecular phenolic acids; Cluster VII, amino acids; Cluster VIII, saccharides, respectively.The absolute value of the Pearson correlation coefficient is between 0-1, ≥ 0.8 is highly correlated, 0.5–0.8 is moderately correlated, 0.3–0.5 is lowly correlated, and < 0.3 the correlation is extremely weak and can be regarded as irrelevant.

In Fig.9, we found most of flavonol glycosides were not correlated to the bitter and astringent scores of tea infusion after roasting.It suggested that the changes of roasted tea taste (bitterness and astringency) did not rely on the low-threshold flavonol glycosides.Meanwhile, hydrolysable and condensed tannins had high Pearson correlation coefficients (≥ 0.8) with bitterness and astringency, andN-ethyl-2-pyrrolidione-substituted flavan-3-ols may be effective in decreasing the bitter and astringent perceptions because of the negative Pearson correlation coefficients [29].

4.Conclusion

Roasting changes the chemical composition of tea, and then affect the taste perceptions on bitterness and astringency.In the present study, the main differences between non-roasted teas and roasted teas were the content of flavan-3-ols and GAs.The low-astringent and bitter perceptions should be ascribed to the lower concentrations of these astringent compounds.Furthermore, the stimulated roasting test also provided more detailed information about the relationship between chemical composition and bitter/astringent taste.It revealed that theN-ethyl-2-pyrrolidione-substituted flavan-3-ols, and coumaroylquinic acid were tightly correlated to the taste variation.

Conflict of interest

The authors declare no competing financial interest.

Acknowledgements

This work was supported by Natural Science Foundation of China (32072633, 32072634, 31201335) , earmarked fund for China Agriculture ResearchSystem (CARS-19), Anhui Key research and development plan (1804b06020367, 202004b11020004) and Young Elite Scientist Sponsorship Program by National CAST(2016QNRC001).