PCA of five different grades of stir-fried green tea
To minimize the influences of other factors, samples from each stage were collected from the same producer on the same day. Supplemental Fig. 1 and Supplemental Fig. 2 present the total ion chromatograms of the quality control sample and substances detected during all processing stages of stir-fried teas. In total, 898 metabolites were detected in the samples and profiled in a heatmap (Fig. 2a). The relative contents of individual compounds were similar between the three biological duplicates of each stir-fried tea but remarkably different among the different stages of stir-fried tea production, except the stir-fried metabolites between S3 and S4.
Heatmap and principal component analysis (PCA) plot of three Yuhua tea samples with three duplicates. (a) Heatmap; (b) 2D score scatter plots of the PCA for five samples (S1, S2, S3, S4, S5 and QC). The QC (quality control) sample was a mixture of S1, S2, S3, S4 and S5.
To visualize the sample similarities, the metabolites were subjected to a PCA (Fig. 2b) to preliminarily understand the overall metabolic differences between samples in different groups and the variations among samples within groups. Among groups, there was a trend of metabolomics separation, indicating differences in metabolomics between sample groups. The mixed quality control samples clustered in the center of the PCA score plot, indicating that the compound extractions and LC–MS analysis were reliable.
The first and second PCs explained 32.25% and 15.31% of the total variance, respectively. In the PCA score plot, the S3 and S4 samples were closely clustered, and there were significant separations among S1, S2, S3/S4 and S5. As a result, S1 and S5 were distinctly separated only by PC1 (32.25%), S1 and S2 were clearly separated by PC2 (15.31%) and they were also discriminated by PC1 (32.25%). The other samples were significantly separated by PC1 (32.25%) and PC2 (15.31%), with the exception of S3 and S4.
The PCA score plot showed that the metabolites at each stage had changed significantly, except S3 to S4. S3 was close to S4 on the PCA score plot, suggesting that the metabolites of S3 were similar to those of S4, indicating that spreading, fixation and drying caused the metabolites of tea leaves to change distinctly.
Changes in the main biochemical component content in tea during different manufacturing processes
The main biochemical components of green tea are tea polyphenols, amino acids and caffeine. The content and ratio of each component determines the quality of the tea's taste. The astringency and bitterness of green tea are believed to be closely related to some nonvolatile compounds, including polyphenols, catechins purine alkaloids and caffeine. Thus, in this study, we detected the content of tea polyphenols, caffeine and free amino acids using UPLC (Fig. 3). The tea polyphenol content increased significantly during the fixation and drying processes but decreased obviously during the rolling stage (Fig. 3a). The differences in caffeine content were not statistically significant among the five manufacturing stages (Fig. 3b). Amino acids determine the fresh taste of green tea, and, although the amino acid content showed an obvious downward trend, the content in the final production stage remained high (Fig. 3c).
Changes in the main biochemical component content of tea during different manufacturing processes. The changes in (a) tea polyphenols, (b) caffeine and (c) free amino acid content during different tea processing stages.
Characterization of four comparisons of differential metabolites
A total of 898 compounds were extracted, and OPLS-DA modeling was applied to the LC–MS datasets to determine which metabolites were significantly changed during the different stage of Yuhua green tea manufacturing (Fig. 4a–e). Compared with a PCA, OPLS-DA maximizes the differentiations between groups, which is helpful for finding differential metabolites. OPLS-DA combines orthogonal signal correction and PLS-DA, which decomposes the X-matrix information into Y-related and irrelevant information and selects the difference variables by removing the irrelevant differences. Using the OPLS-DA model, the metabolomics data were analyzed, and a score map of each group was constructed to further illustrate the differences among the groups. These models were constructed with data from samples taken at different stages of the tea manufacturing process. Potential biomarkers for separation by shade effects were subsequently identified using S-plots, which represented covariance (p) against correlation (p-corr).
Orthogonal projections to latent structures-discriminant analysis (OPLS-DA). Score scatter plots of the OPLS-DA model for S1 vs S2 (a), S2 vs S3 (b), S3 vs S4 (c), S4 vs S5 (d) and S1 vs S5 (e).
Pairwise comparisons of the metabolite contributions between S1 and S2, S2 and S3, S3 and S4, and S4 and S5 were performed using the OPLS-DA models (Supplemental Fig. 3) further supporting that the two models were meaningful, and their differential metabolites were screened using the VIP value analysis. As shown in Fig. 3 and Fig. 4, each comparison was clearly separated, with the S1 triplicates all grouped to the left and the S2 triplicates to the right (Fig. 4a). Similar results were obtained between S1 and S2, S2 and S3, S3 and S4, S4 and S5, and S1 and S5 (Fig. 4b–e), providing visual evidence of the clustering of tea samples by manufacturing process.
The above OPLS-DA models were then used to construct an S-plot (Fig. 5a–e), which provided a graphical projection of specific compounds. Metabolites far from the plot origin contributed greatly to the separation between different samples. The abscissa represents the covariance between the PC and the metabolite, and the ordinate represents the correlation coefficient between the PC and the metabolite. The closer the metabolite is to the upper right corner and the lower left corner, the more significant the difference. The red dots indicate that the VIP values of these metabolites are greater than or equal to 1, and the green dots indicate that the VIP values of these metabolites are less than 1. As shown in Fig. 5a and 5b, 167 and 150 compounds (Supplemental Table 2) contributed significantly to the separations between S1 and S2 and between S2 and S3, respectively. As shown in Fig. 5c and 5d, 26 and 95 compounds (Supplemental Table 2) are significantly different metabolites that contribute to the separations between S3 and S4 and between S4 and S5, respectively. As shown in Fig. 5e, 306 compounds (Supplemental Table 2) are significantly different metabolites that contribute to the separations between S1 and S5. They are the compounds from the S-plot that were farthest in the positive and negative directions, respectively, from the origin. The farther a compound was from the origin, the greater its contribution to the distinction between samples.
Orthogonal projections to latent structures-discriminant analysis (OPLS-DA). S-plots of the OPLS-DA model for S1 vs S2 (a), S2 vs S3 (b), S3 vs S4 (c) and S4 vs S5 (d), S1 vs S5 (e).
The heatmap analysis characterized the relative content of the recognized differential metabolites in the Yuhua teas at the five different tested stages (Fig. 6). The color coding, from red to green, indicates their relative content, from high to low, respectively. As shown in Fig. 6a–d, the differential metabolites were clearly clustered into two differently colored sections, indicating that significant differences existed between S1 and S2, S2 and S3, S3 and S4, and S4 and S5.
Four pairwise heatmaps of all the identified compounds from the five stir-fried green teas. Pairwise heatmaps of the relative differences in metabolites between (a) S1 and S2, (b) S2 and S3, (c) S3 and S4, (d) S4 and S5 and (e) S1 vs S5. Red and green indicate higher and lower abundances, respectively.
Changes in the metabolites between S1 and S2
As shown in Fig. 7a, 167 significantly different metabolites were detected between S1 and S2. The expression levels of 116 metabolites were obviously increased and 51 metabolites were clearly decreased. Flavonoids, organic acids and their derivatives, amino acids and their derivatives, and lipids were mainly increased metabolites, including engeletin, apigenin C-glucoside, quercetin 3-O-rutinoside, quercetin 7-O-rutinoside, isovitexin 7-O-glucoside, trans-citridic acid, sebacate, punicic acid, proline, quercetin O-acetylhexoside and luteolin 3',7-di-O-glucoside. The flavonoids (47 kinds) accounted for the largest proportion among the significantly increased metabolites. The decreased metabolites mainly included nucleotides and their derivatives, isoacteoside, and vitamins and their derivatives, including p-coumaraldehyde, hydrocinnamic acid, L-ascorbate, niacinamide and 2-(formylamino) benzoic acid.
Differential metabolite analyses for different stages using the criteria VIP ≥ 1 and FC ≥ 2 or ≤ 0.5. Volcano plots of differential metabolites between (a) S1 vs S2, (b) S2 vs S3, (c) S3 vs S4, (d) S4 vs S5 and (e) S1 vs S5. Red and green dots represent increased and decreased differential metabolites, respectively; gray dots represent non-differential metabolites.
As the main secondary metabolites in plants, flavonoids are the most important quality-related compounds, contributing to the color, taste and aroma of brewed tea. The synthesis and accumulation of flavonoids occurs in response to environmental cues. Flavonoids are the main growth and defense regulators in plants, and they are induced and biosynthesized as the result of long-term natural selection and acclimatization processes. The spreading of freshly plucked leaves may represent a stressful environmental change, which results in the increased metabolic levels of flavonoids. In addition, the proline content increased, which also indicates a stress response. Moreover, during this process, the overall expression levels of amino acids and their derivatives increased significantly, which may be as a result of protein degradation initiated by proteases and peptidase released from damaged cells or dissimilated from sugars.
Changes in the metabolites between S2 and S3
In total, 150 significant differences in metabolites were found between the S2 and S3 samples. In the score plots (S2 vs S3, Fig. 2b), S2 and S3 are obviously separated, indicating that during fixation, the metabolites have changed significantly. Among these metabolites, 98 and 52 increased and decreased, respectively (Fig. 7b). Lipids represented the most increased metabolites. The number of nucleotides and their derivatives increased by 16, slightly less than that of lipids (19). In addition, there were 15 flavonoids, 10 phenylpropanoids, seven organic acids and derivatives, six amino acid and derivatives, four vitamins and derivatives, three carbohydrates and other metabolites that increased obviously. In addition, 11 flavonoids, nine phenylpropanoids, eight organic acids and derivatives, six lipids, three amino acid and derivatives, two nucleotide and derivatives and other kinds of metabolites decreased. As shown in Supplemental Table 1, oxoadipic acid, trans-citridic acid, C-hexosyl-isorhamnetin O-hexosi, luteolin 3',7-di-O-glucoside, procyanidin A3, apigeninidin chloride, bilobalide, tricin O-feruloylhexoside O-hexoside, C-hexosyl-luteolin O-feruloylpentoside and maslinic acid were the most significantly decreased metabolites.
Thus, fixation is the critical time period for the increase in lipids involved in the flavor quality of tea. Lipid-soluble chlorophylls and carotenoids are the main pigments in tea plants, and they impact the color quality of tea products, especially green tea. During the production of green tea, especially during fixation, significant lipidomic variations were observed, which were mainly related to chlorophyll decomposition, phospholipid acid reduction and glycolipid degradation, which may contribute to the color and aroma qualities of tea. The fixation stage is the key period of chlorophyll decomposition and pheophytin production; therefore, these results are consistent with those of previous studies[31,33].
Changes in the metabolites between S3 and S4
Of the 26 differentially expressed metabolites between S3 and S4 (Fig. 7c), ten were up-regulated and 16 were decreased. Except for organic acids, amino acids, flavonoids and phenylpropanoids, the number of metabolites decreased more than increased. Thiamine, luteolin 3',7-di-O-glucoside, C-hexosyl-isorhamnetin O-hexoside, C-hexosyl-luteolin O-feruloylpentoside, cocamidopropyl betaine, pantothenol, 8-hydroxyguanosine, tangeretin, oxalic acid and docosanoic acid decreased significantly, while 4-O-caffeoyl quinic acid, chlorogenic acid methyl ester, 2-(formylamino) benzoic acid, 6-hydroxydaidzein, engeletin, myricitrin, kaempferin, tricin 7-O-hexosyl-O-hexoside, tricin 7-O-β-guaiacylglycerol, 5-methylcytosine, coumestrol, angelicin and γ-Glu-Cys increased obviously.
Changes in the metabolites between S4 and S5
Of the 95 differentially expressed metabolites between S4 and S5 (Fig. 7d), 58 were increased and 37 were decreased. During the drying process, the change in flavonoids was the most significant, with 22 increasing and eight decreasing. Among the phenylpropanoids, four increased and 11 decreased. Additionally, there were five increased and one decreased organic acid and their derivatives during this process. Heat treatments may induce the release of natural organic acids, resulting in increased acidity. Therefore, during the drying process, the increase in the organic acids and derivatives may be due to the heat treatment. Additionally, five kinds of lipids showed either significant increases or no significant reductions during this process. Under heat–acid stress, chlorophyll is destroyed and lipid levels increase.
Changes in the metabolites between S1 and S5
In the comparison of each stage, the metabolites with significant changes between S1 and S5 were the most, regardless of whether they decreased or increased. The expression levels of 210 metabolites were significantly increased and 96 metabolites were clearly decreased. The significantly increased metabolites are mainly concentrated in amino acids and derivatives, flavonoids and lipids, and metabolites such as organic acids and derivatives, nucleotide and derivatives, anthocyanins and phenylpropanoids mainly decreased.