Tea shoots were harvested from the tea plant Camellia sinensis (L.) Jinguanyin to obtain disease-free samples with one bud and three leaves in April 2021. The tea plants were cultivated in an artificial tea plantation in Anxi County, Fujian Province (24.55°N, 117.50°E), China. JGY was extensively planted and processed into oolong tea because of its excellent qualities and high economic efficiency.

The tea was manufactured according to prior research with some modifications to achieve the flavor characteristics of Anxi Tieguanyin (4). The manufacturing processes of oolong tea and the sampling points used in this study are described as follows (Figure 1). The withering of fresh leaves (WT) was performed by placing them indoors for a three-hour resting period and flipping between them twice. Then, tea samples were turned over in a rotary machine, followed by a setting step of standing and spreading out on bamboo sieves, where the steps of turning-over and setting were performed three times. All setting steps were carried out in an air-conditioned room (temperature 20°C, relative humidity 60%), and the relevant parameters are presented in Table 1. Immediately thereafter, the tea samples were subjected to a firing treatment with a rotary drum dehydrating machine for 3 min at 270°C. Finally, the mixtures were rolled at room temperature (26°C–28°C) and dried at 65°C to yield the tea product (PT). The samples included fresh leaf (FL), after withering leaf (WT), after first turning-over leaf (FT), after second turning-over leaf (ST), after third turning-over leaf (TT), before firing leaf (BF), and tea product (PT). Three biological replicates for each sample point were obtained, immediately frozen and fixed in liquid nitrogen, and then stored in a −80°C refrigerator for study.

Electronic tongue (TS-5000Z, Insent Electricity Company, Japan) was used to assess tasting attributes of finished tea, and this machine which has six basic taste sensors contributed to evaluation of various tasting attributes. JGY tea product (3.0 g) was brewed of 150 mL boiling water for 5 min, then rapidly filtered and cooled to indoor temperature. Finally, the filtered tea infusion was detected and repeated three times. Simultaneously, the mixed liquids of 30 mM potassium chloride and 0.3 mM tartaric acid were used as reference solution and as pre-treatment solution of taste sensors (tasteless solution).

JGY tea product was performed to sensory evaluation according to previous studies (15, 16), panelists (five females and five males, 20 to 50 years old) had more than 5 years of sensory evaluation experience and were trained. Brewing was conducted according to the review method with cylindrical cups in the methodology for sensory evaluation of tea (GB/T 23776-2018), 3.0 g finished tea was immersed for 5 min with boiling water. Then intensity (0–10) of tasting attributes (mellow, umami, astringency, bitterness, and thick) was recorded and scored, and different intensities of taste attributes represented by 0 to 10 were also used in previous studies (17).

The metabolites of all samples in this experiment were extracted according to previous studies (18). Specifically, frozen samples were freeze-dried using a vacuum freeze-dryer (Scientz-100F) and crushed at 30 Hz for 1.5 min using a mixer mill (MM400, Retsch) with zirconia beads. One hundred milligrams of lyophilized powder was solubilized in 1.2 mL of 70% methanol solution, rotated every 30 min for 30 s six times, and left overnight in a refrigerator at 4°C. The samples were centrifuged at 12,000 rpm for 10 min and filtered prior to UPLC-MS/MS analysis (SCAA-104, 0.22 μm pore size; ANPEL, Shanghai, China, http://www.anpel.com.cn/). The quality control sample was a mixture of all materials (mix) in equally divided masses, which was used to ensure the stability and reproducibility of the experiment.

A UPLC-ESI-MS/MS system (UPLC, SHIMADZU NexeraX2, www.shimadzu.com.cn/; MS, Applied Biosystems 4,500 Q TRAP, www.appliedbiosystems.com.cn/) was used to analyze these example extracts. The analysis was conducted with a UPLC column (Agilent SB-C18, 1.8 μm, 2.1 mm × 100 mm). Mobile phases A and B were pure water with 0.1% formic acid and acetonitrile with 0.1% formic acid, respectively. Sample measurements were obtained using a gradient program that employed the starting conditions of 95% A and 5% B. A linear gradient was programmed to 5% A and 95% B over 9 min, and the 5% A and 95% B composition was maintained for 1 min. Afterward, a 95% A and 5.0% B composition was reached within 1.10 min and maintained for 2.9 min. The flow rate was 0.35 mL/min, the column oven temperature was 40°C, and the injection volume was 4 μL.

Eluate connection to a triple quadrupole-linear ion trap mass spectrometer (Q TRAP) in an AB4500 Q TRAP UPLC/MS/MS system with an ESI Turbo Ion-Spray interface, operating in positive and negative ion modes under the control of Analyst 1.6.3 software (AB Sciex), was used to perform LIT and triple quadrupole (QQQ) scanning, simultaneous instrument tuning and mass calibration with 10 and 100 μmol/L solutions of polypropylene diol. Qualitative and quantitative profiling of metabolites was conducted based on retention time (RT), fragment patterns, accurate m/z values and MWBD (home-made database, Metwwere, Wuhan, China) (18). Quantitation of metabolites was accomplished using triple quadrupole mass spectrometry in multiple reaction detection mode (MRM). Upon acquiring metabolite spectra from various samples, the peaks in the mass spectra were integrated for all compounds, and integration correction was performed for the same substance between samples (19).

The experiment was conducted in triplicate, and the average of the three replicate peaks was used for analysis. Various statistical analysis methods were applied to the metabolite data, including T testing, principal component analysis (PCA), orthogonal partial least squares-discriminant analysis (OPLS-DA), hierarchical cluster analysis (HCA), and K-means clustering. T tests were conducted to evaluate statistical significance, and OPLS-AD and fold change (FC) were used to screen for significant differences in metabolites between treatments, with the screening standards of VIP ≥1 and FC ≥1.6 for significantly upregulated differential compounds and VIP ≥1 and FC ≤0.62 for significantly downregulated differential compounds (20). K-means clustering analysis was conducted to analyze the dynamics of metabolites between treatments, along with metabolite enrichment versus annotation analysis using the KEGG database¹ (21), and all algorithms and mappings were implemented using the R package.

The listed chemicals were all HPLC grade. Methanol, acetonitrile, formic acid, polypropylene glycol, and DMSO were purchased from Merck Chemical Technology Co., Ltd. (Shanghai, China).

The combination of electronic tongue (ET) and sensory evaluation make the review of tea more objective and scientific. Firstly, tasteless solution was used as the control experiment of ET (Figure 2A), sourness and saltiness were excluded owing to less than tasteless values, and other higher tasting properties were shown in Figure 2B. The difference between two points (sweetness and astringency) and tasteless values was the largest, followed by umami, aftertaste-A, and bitterness. The result of sensory evaluation was investigated in Table 2, quantitative description test was shown in Figure 2C. In summary, JGY tea products present a characteristic flavor of mellow, sweet in aftertaste, and significant Yin rhyme.

A total of 1,005 metabolites were detected and classified into 12 major categories, including 87 amino acids and derivatives, 154 phenolic acids, 57 nucleotides and derivatives, 249 flavonoids, 30 lignans and coumarins, 29 tannins, 71 alkaloids, 133 lipids, 62 saccharides and alcohols, and 35 other components (Figure 3A). The peak areas of all metabolites and their dynamic changes during manufacturing were visualized with clustering heatmaps (Figure 3B). The results showed that FL, WT, FT and ST were clustered into one group and TT, BF and PT were clustered into another group, which indicated that some metabolites of the tea leaves presented similar trends at FL, WT, FT and ST, while others changed similarly at TT, BF and PT. Many metabolites accumulated abundantly at TT, BF and PT, suggesting that the third turning-over, third setting and high-temperature treatments could be critical elements for accumulation in some compounds. Meanwhile, the three biological replicates of each sample were aggregated together, reflecting excellent stability and reproducibility within groups.

The principal component analysis results for the tea manufacturing and quality control (mix) samples are shown in Figure 3C, with the top three principal components cumulatively contributing 60.46% (PC1 = 35.26%, PC2 = 16.96% and PC3 = 8.24%). The quality control samples (mix) were tightly clustered together and positioned in the core among all samples, indicating greater dependability and duplicability of the resulting data. The PCA plot of the manufacturing samples showed great separation between various treatment samples and changes in the metabolite fractions with the manufacturing of oolong tea. Interestingly, there was a relatively large difference between FL WT FT ST and TT BF PT in the direction of PC1, suggesting that the tea quality changes in oolong tea manufacturing were strongly related to PC1 components. This result was consistent with the results shown in the clustering heatmap.

The changes in the nonvolatile contents of diverse types in oolong tea manufacturing were calculated, as shown in Figure 3D. According to the results, the samples were arranged in terms of the total amount of nonvolatile metabolites in the following sequence: PT > TT > BF > ST > FT > WT > FL. Moreover, 9 metabolites (all studied metabolites except phenolic acid, lignans and coumarins) changed significantly during manufacturing (p < 0.05). Among them, the accumulation of amino acids and their derivatives and alkaloids increased constantly from FL to BF but decreased with the high-temperature treatment (PT). Phenolic acids, organic acids, tannins, and lipids achieved maximum abundance when the third turning-over ended or manufacturing was completed. Nucleotides and their derivatives displayed a sustained upward trend, while flavonoids, sugars and alcohols displayed a negative trend. Collectively, amino acids and their derivatives, alkaloids, flavonoids, sugars, organic acids and lipids underwent remarkable alterations during the manufacturing of oolong tea.

Previous research has demonstrated that the origin of amino acids during the manufacturing of oolong tea lies mainly in the degradation of proteins (22, 23); meanwhile, several soluble sugars were transformed into aromatic chemicals, such as pyrazines and pyrroles, via the Maillard reaction with amino acids under the influence of heat. The dynamic changes in amino acids and their derivatives in the current research were probably related to protein hydrolysis before the high temperature treatment and the massive occurrence of Maillard reactions between amino and carboxy compounds at high temperature. Flavonoids are key contributors to the flavor profile of tea and precursors to important metabolites, such as flavanols (catechins) and flavonoid (alcohol) glycosides (24). There was an overall consistent decreasing trend of flavonoids in oolong tea manufacturing, which indicates that flavonoids were gradually converted to other compounds, causing a decrease in their content.

To explore the influence of the various manufacturing processes of oolong tea on diverse compounds, FL vs. WT, WT vs. FT, FT vs. ST, ST vs. TT, TT vs. BF, BF vs. PT, and FL vs. PT were compared for differential metabolite identification and analysis. The overall numbers of differential metabolites in each grouping are shown in Figure 4A, which revealed the following trend of differential metabolite numbers: TT > BF > PT > ST > WT > FT. This trend indicated that the greatest extent of metabolite variation in tea leaves was between the second turning-over and third turning-over (257), followed by TT vs. BF (206) and BF vs. PT (175), while the withered leaves (112), the first turning-over leaves (101) and the second turning-over leaves (115) were comparatively close, which was in agreement with the results of PCA and clustering heatmap analysis. Interestingly, more compounds increased in content than decreased in all groups (Figure 4B).

Common and characteristic differential metabolites in different groups were identified with Venn diagrams (Figure 4C). More specifically, six differential metabolites commonly associated with six groups (in addition to FL vs. PT) were screened, including 6-hydroxy-5,7,4′-trimethoxyflavone, 6-hydroxy-5,7,4′-trimethoxyflavone, lysoPE 16:0, lysoPC 16:2 and benzylacetone, three of which belonged to the flavonoid group and exhibited identical trends, and all reached their peak at point BF, which was possibly attributable to the oxidation of flavonoids. There were 7, 4 and 13 unique differential metabolites affected by the withering, first turning-over and second turning-over processes, respectively, while 76, 56 and 63 differential metabolites specific to the ST vs. TT, TT vs. BF and BF vs. PT groups, respectively, accounted for 89.04% of the total unique differential metabolites (219).

Hence, some compounds were absent or present in low amounts in fresh leaves of tea, but they were obtained or their contents were increased through oolong tea manufacturing (withering, the first turning-over, the second turning-over, the third turning-over, setting quietly and high-temperature treatment), and the third turning-over, setting and high-temperature treatment were the most key moments of metabolite changes in oolong tea processes.

To investigate the relative content change trends of metabolites in different groups, differential metabolites that were identified according to the screening criteria in all comparative groups were subjected to z-score normalization followed by K-means clustering analysis.

A total of 562 differential metabolites were clustered into 12 clusters, with variation trends falling into three categories (Figure 5). Among them, the compounds of clusters 2, 3 and 12 exhibited increasing trends during oolong tea manufacturing, which mainly included indole, L-phenylalanine, L-glutamine, and dihydrokaempferol, and large accumulations of these compounds were conducive to the formation of oolong tea characteristics. Second, the compounds of clusters 1 and 11 tended to decrease in content due to being consumed, such as D-glucose, D-fructose, and 2-methoxycinnamic acid, which are important sources of secondary metabolism in oolong tea. Finally, the compounds of seven clusters (4, 5, 6, 7, 8, 9 and 10) showed fluctuating variations, with all peaking at TT or BF. In more detail, the compounds of clusters 4, 5 and 9 reached their highest peaks at TT, which consisted of γ-aminobutyric acid, jasmonic acid, xanthosine, myricetin, and citric acid; this indicates that turning over promoted the accumulation and biosynthesis of these compounds. Interestingly, compounds of clusters 11 and 12 were dramatically changed after high-temperature treatment, including naringenin, kaempferol, and naringenin chalcone, suggesting that high-temperature treatment in oolong tea manufacturing is a critical influencing factor for these compounds.

A total of 562 differential metabolites were classified as shown in Supplementary Figure S1, mainly including 59 amino acids and their derivatives, 91 phenolic acids, 33 nucleotides and their derivatives, 93 flavonoids, 23 lignans and coumarins, 20 tannins, 43 alkaloids, 39 organic acids, 101 lipids, 28 saccharides and alcohols, and 18 others.