Untargeted metabolomics and targeted quantification revealing polyphenol remodeling and gallic acid enrichment during storage of Wuyi rock tea Rougui

Abstract

Wuyi rock tea Rougui samples produced consecutively from 2016 to 2025 were analyzed using UHPLC-QTOF-MS untargeted metabolomics combined with UHPLC-QqQ-MS/MS targeted quantification to characterize polyphenol variation. 17 catechin-gallic acid-related annotated features were retained, suggesting that storage-associated variation involved a broad range of polyphenolic constituents. Targeted quantification of eight representative catechins and gallic acid revealed that galloylated and non-galloylated catechins varied in trend and magnitude across years, with compositional redistribution between the two fractions and without a consistent increase or decrease. Gallic acid content was higher in the samples with longer storage time, with increases of 166.25% and 168.25% in RG16 and RG17 compared with RG25, respectively. Correlation analysis indicated positive associations between EGCG–GCG and ECG–CG, while GA showed near-zero linear correlations with EGCG and GCG. These findings suggest that long-term storage of Rougui is associated with polyphenol compositional remodeling involving catechin redistribution and gallic acid enrichment.

Highlights

• •Same-origin Rougui rock tea samples from ten consecutive production years were analyzed by untargeted and targeted LC–MS profiling.
• •Seventeen catechin–gallic acid-related polyphenols were annotated from the untargeted metabolomic dataset.
• •Targeted quantification revealed redistribution between galloylated and non-galloylated fractions and an increase in gallic acid as storage time increased.
• •Correlation analysis revealed coordinated variation among targeted polyphenols, while gallic acid showed near-zero linear correlations with the major galloylated compounds.

1. Introduction

The quality value of aged Wuyi rock tea is closely associated with the gradual transformation of non-volatile constituents during long-term storage. Tea storage involves multiple biological, chemical, and physical processes, and the mechanisms underlying these quality changes remain incompletely understood (Lv et al., 2023). Studies on stored teas have shown that storage duration can substantially affect non-volatile metabolic profiles, with associated changes in polyphenols, flavonoids, organic acids, amino acids, and their derivatives. For instance, investigations of black tea and white tea storage have indicated that long-term storage induces progressive changes in both metabolite composition and quality-related constituents (Huang et al., 2021; Zhao et al., 2022), while characteristic catechin derivatives have been used to discriminate storage duration in white tea metabolomics (Dai et al., 2018). Raw Pu-erh storage studies have similarly demonstrated that storage time and environmental conditions influence metabolite composition and taste quality (Xu et al., 2019). Oolong tea storage has also been reported to alter non-volatile constituents including catechins, gallic acid, amino acids, and lipids (Peng et al., 2022). For Wuyi rock tea, a metabolomics study of Shuixian across different storage years identified five compounds positively correlated with storage duration (Song et al., 2024). Targeted analyses have quantitatively identified gallic acid, representative catechin monomers, and theaflavins as major non-volatile markers (Ai et al., 2024). Nevertheless, a systematic approach linking global metabolic variation in continuously produced year-series samples to specific non-volatile constituents is still lacking. Previous metabolomics studies suggest that tea post-processing involves broader non-volatile metabolic remodeling beyond a limited set of marker compounds (Long et al., 2020). More broadly, temporal and environmental factors can shape the chemical and functional characteristics of natural systems over time (Bera & Nag, 2025; Hu et al., 2025). However, whether similar large-scale metabolic variation occurs during long-term storage of Wuyi rock tea Rougui remains to be elucidated.

To address this issue, Rougui samples from a single origin covering 10 consecutive production years (2016–2025) were analyzed using a dual LC-MS platform comprising UHPLC-QTOF-MS untargeted metabolomics and UHPLC-QqQ-MS/MS targeted quantification. UHPLC-QTOF-MS was employed to characterize overall metabolic differences across production years and to screen for storage-associated annotated features, while UHPLC-QqQ-MS/MS was used for the absolute quantification of representative polyphenols indicated by the untargeted results. By combining global untargeted screening with targeted quantitative validation, this study aimed to identify chemically interpretable constituent groups underlying the overall metabolic variation during Rougui storage, thereby providing a basis for understanding the chemical changes and quality assessment of aged rock tea.

2. Materials and methods

2.1. Samples and pretreatment

Rougui (Camellia sinensis) samples (RG) were provided by Youlong Rock Tea Factory, Wuyishan, Fujian, China. Fresh leaves were harvested from Mituo Rock tea garden (27.703°N, 117.972°E), and processed according to traditional Wuyi rock tea manufacturing procedures (Zheng et al., 2025). All dried tea samples were aged in sealed aluminum foil bags under controlled storage conditions at 25 °C with relative humidity below 50%. The sample set comprised 10 consecutive production years from 2016 to 2025, designated RG16 through RG25, corresponding to storage durations of 9 to 0 years, respectively. Three biological replicates were prepared for each production year.

For sample preparation, dried tea was ground and sieved through a 60-mesh (250 μm) screen. Approximately 0.1 g of tea powder was extracted with 10 mL of 70% aqueous methanol by ultrasonic extraction for 30 min under ice-water bath conditions (T ≤ 10 °C), using a KQ-5200DE ultrasonic cleaner (Kunshan Ultrasonic Instruments Co., Ltd., Jiangsu, China) at 200 W and 40 kHz. The extract was centrifuged at 12,000 rpm for 15 min at 4 °C (H3-16KR refrigerated centrifuge; Kecheng Instrument Equipment Co., Ltd., Hunan, China), and the supernatant was collected and subjected to a second centrifugation under the same conditions. The clarified supernatant was filtered through a 0.22 μm PTFE membrane, and the filtrate was diluted 100-fold with 70% aqueous methanol before LC-MS analysis.

2.2. Reagents and standards

(+)-Catechin (C), (−)-epicatechin (EC), (−)-catechin gallate (CG), (−)-epicatechin gallate (ECG), gallocatechin (GC), (−)-epigallocatechin (EGC), gallocatechin gallate (GCG), (−)-epigallocatechin gallate (EGCG), and gallic acid (GA) (all HPLC ≥98%) were purchased from Beijing Solarbio Science & Technology Co., Ltd. (Beijing, China). Methanol, acetonitrile, and formic acid (all LC-MS grade, ≥99.99%) were obtained from Macklin Biochemical Co., Ltd. (Shanghai, China). Ultrapure water was prepared with a Milli-Q purification system (Millipore, Billerica, MA, USA). Individual stock solutions of catechin and gallic acid standards were prepared in 70% aqueous methanol, stored at −20 °C, and freshly diluted with 70% aqueous methanol to prepare mixed working solutions and calibration standards before analysis.

2.3. Untargeted global metabolic profiling by UHPLC-QTOF-MS

Untargeted metabolomic profiling was performed on an LC-30A UHPLC system (Shimadzu, Kyoto, Japan) coupled to a TripleTOF 6600+ high-resolution mass spectrometer (SCIEX, Foster City, CA, USA). Chromatographic separation was achieved on a Waters ACQUITY UPLC HSS T3 column (100 mm × 2.1 mm, 1.8 μm) at 40 °C. Mobile phase A was 0.1% formic acid in ultrapure water, and mobile phase B was 0.1% formic acid in acetonitrile. The flow rate was 0.4 mL/min and the injection volume was 4 μL. Gradient elution was programmed as follows: 0–5 min, 5%–65% B; 5–6 min, 65%–99% B; 6–7.5 min, 99% B; returning to 5% B at 7.6 min and re-equilibrating to 10 min. Chromatographic and MS conditions were adapted from a published untargeted metabolomics study of Wuyi rock tea across storage years (Song et al., 2024) and further optimized for the present samples. Detailed MS acquisition times, ion source parameters, scan ranges, and collision energy settings are provided in Table S1.

2.4. Targeted quantification by UHPLC-QqQ-MS/MS

Targeted quantification was performed on a Nexera XS UHPLC system coupled to an LCMS-8050 triple quadrupole mass spectrometer (Shimadzu, Kyoto, Japan). Chromatographic separation was carried out on a Shim-pack GIST-HP C18 column (100 mm × 2.1 mm, 3 μm) at 39 °C. Mobile phase A was 0.1% formic acid in ultrapure water, and mobile phase B was 0.1% formic acid in acetonitrile. The flow rate was 0.20 mL/min, the injection volume was 5 μL, and solvent compressibility coefficients for pumps A and B were set to 0.45/GPa and 1.20/GPa, respectively. The linear gradient was: 0 min, 10% B; 6.15 min, 30% B; 8.80 min, 70% B; 9.10 min, 80% B, held to 11.00 min; returning to 10% B at 11.10 min with re-equilibration to 15.00 min.

Detection was performed using an electrospray ionization (ESI) source in positive/negative ion switching mode. The eight representative catechins were detected in positive ion mode, and GA was detected in negative ion mode. Multiple reaction monitoring (MRM) transitions, collision energies, and overall MS interface parameters are listed in Table 1 and Table S2; additional parameters including Q1/Q3 pre-bias voltages and dwell times were optimized by the auto-tuning function of LabSolutions software. Chromatographic resolution of structurally related catechin pairs was evaluated, and RT, FWHM, and Rs values are provided in Table S4. The method was adapted from published UHPLC-MS/MS procedures for tea catechins and related non-volatile constituents (Jiang et al., 2019; Tao et al., 2016; Wang et al., 2022) and further optimized according to the chromatographic behavior and signal responses of the present Rougui samples.

Table 1.

Compound-dependent MRM parameters for UHPLC-QqQ-MS/MS analysis of catechins and gallic acid.

Type	Analyte	Precursor ion (m/z)	Quantifier ion (m/z)	Qualifier ion (m/z)	CE Quantifier (eV)	CE Qualifier (eV)	Polarity
MRM	C	291.2	139	123	−16	−22	POS
MRM	GC	307.2	139	163	−18	−22	POS
MRM	EGC	307.2	139	163	−35	−35	POS
MRM	EGCG	459.2	139	289.1	−18	−28	POS
MRM	EC	291.2	139	123	−14	−16	POS
MRM	GCG	459.2	139	289.1	−18	−28	POS
MRM	ECG	443.2	139	123	−22	−26	POS
MRM	CG	443.2	139	123	−22	−26	POS
MRM	GA	169.05	125.05	79.1	10	24	NEG

2.5. Data processing, multivariate analysis, targeted statistics, and method validation

Raw UHPLC-QTOF-MS data (.wiff) were converted to .mzML format using ProteoWizard and processed in R (v4.5.2) within RStudio (Posit team, Boston, MA, USA). Peak detection, retention time correction, and peak alignment were performed with the XCMS package (v4.8.0). Features with a missing rate > 50% across samples were excluded. Remaining missing values were handled using a two-step imputation strategy: features absent in >50% of blank samples were replaced with one-fifth of the minimum observed value, while other missing values were imputed by k-nearest neighbor algorithm (KNN; impute package). Peak area correction was applied using support vector regression (SVR; e1071, v1.7–16). Metabolite annotation was performed by matching against the METWARE commercial database using the metDNA algorithm; features with a composite annotation score > 0.5 and a QC sample CV < 50% were retained. Positive and negative ion mode datasets were merged and deduplicated to generate the final feature matrix.

PCA (with Hotelling's T² 95% confidence ellipse, including pooled QC samples) and PLS-DA (biological samples only) were performed using the ropls package to characterize overall metabolic separation patterns and storage-associated trajectories. PERMANOVA and PERMDISP were further used to assess overall differences among storage-duration groups and the homogeneity of within-group dispersion, respectively. Both tests were based on Euclidean distances with 9999 permutations, using individual biological samples as the units of analysis; pooled QC samples were excluded. PC1 score trajectories across production years were visualized with ggplot2. Catechin- and gallic acid-related annotated features were further selected from the untargeted dataset, and their year-to-year trajectories were displayed to link global metabolic variation with the subsequent targeted quantification results.

Targeted data were processed in LabSolutions (v5.120; Shimadzu, Kyoto, Japan). Absolute concentrations were derived from seven-point external calibration curves (R² > 0.999) and expressed on a dry-weight basis according to Eq. (1):

$$C_{\mathrm{DW}}\left(\mathrm{mg}/\mathrm{g}\right)=\frac{C_{\text{measured}}\times V_{\text{extract}}\times \mathit{DF}}{m_{\text{sample}}\times 1000}$$ (1)

where C_measured is the extract concentration determined from the calibration curve (μg/mL); V_extract is the extraction volume (mL); DF is the total dilution factor; m_sample is the sample mass (g); and 1000 converts μg to mg. Quantitative trajectories for galloylated catechins, non-galloylated catechins, and GA were plotted using ggplot2. Pearson correlation analysis was performed among the nine targeted compounds, and the correlation matrix was constructed using the GGally, scales, cowplot, and grid packages. Benjamini–Hochberg correction was applied to all 36 unique pairwise correlations, with adjusted P < 0.05 considered significant. Validation of the targeted method followed the ICH Q2(R2) guideline (ICH, 2023).

3. Results and discussion

3.1. Global metabolic variation during long-term storage of Wuyi rock tea Rougui

Untargeted metabolomic profiling of Rougui samples from 2016 to 2025 by UHPLC-QTOF-MS yielded PCA results in which pooled QC samples clustered near the score plot origin (Fig. 1A), confirming satisfactory analytical stability throughout the acquisition. Samples from different production years were distributed across the principal component space, with PC1 and PC2 accounting for 39.3% and 15.2% of total variance, respectively, indicating that storage-associated metabolic variation was predominantly captured along PC1. PLS-DA provided a supervised representation of these storage-related differences (Fig. 1B; R²X(cum) = 0.562, R²Y(cum) = 0.222, Q²(cum) = 0.119). Permutation testing showed that the Q² regression intercept was below zero and the R²Y intercept remained low, with no indication of model overfitting, supporting the validity of the model for visualizing metabolic separation (Fig. 1C).

Fig. 1.

Global metabolic variation of Rougui Wuyi rock tea across production years. (A) PCA score plot of UHPLC-QTOF-MS profiles with pooled QC samples. (B) PLS-DA score plot of biological samples. (C) Permutation validation of the PLS-DA model. (D) PC1 score trajectory across production years. (E) Heatmap of all annotated features after row-wise Z-score normalization. Samples are ordered from RG16 to RG25, with three biological replicates per year.

The PLS-DA model was employed here primarily as a supervised visualization tool to represent storage-associated metabolic structure, rather than as a predictive classifier. The relatively modest Q²(cum) of 0.119 indicates limited predictive performance; however, the supervised separation shown in Fig. 1B remains informative for visualizing structured metabolic variation among the groups. The partial overlap between adjacent production-year groups indicates moderate rather than sharply distinct separation. PCA, PLS-DA with permutation validation, PC1 trajectory analysis, and the global heatmap were therefore used in combination as complementary representations of metabolic variation across the series. PERMANOVA indicated significant overall differences among groups, whereas PERMDISP did not detect significant heterogeneity in within-group dispersion (Table S5). Comparable metabolic separation has been reported in long-term aging studies of dark tea, oolong tea, Wuyi rock tea, and raw Pu-erh tea (Cheng et al., 2021; Hong et al., 2021; Ma et al., 2025; Song et al., 2024).

The PC1 score trajectory across production years revealed that metabolic variation did not follow a single continuous linear trend, but instead showed stage-wise fluctuations (Fig. 1D). PC1 scores were relatively high in RG25 and RG24, decreased notably at RG23, reached lower values in RG22–RG21, and showed a modest partial recovery from RG20 to RG16. The global heatmap of all annotated features further indicated variation across multiple chemical classes (Fig. 1E), suggesting that storage-associated differences were not limited to a few isolated marker compounds but involved concurrent variation across diverse non-volatile constituents. The row-side class annotation in Fig. 1E further showed that this abundance variation was partially organized by chemical class rather than randomly distributed across individual features. Features assigned to related classes, including flavonoids, phenolic acids, and tannins, displayed broadly concordant Z-score patterns across production years, indicating coordinated changes among structurally related non-volatile constituents. This class-level pattern helped narrow the global untargeted variation to chemically interpretable constituent groups. Among these, catechin-, gallate-, and tannin-associated features showed non-monotonic changes, supporting the subsequent focus on the catechins and gallic acid. Because RG16–RG25 represent independently manufactured batches rather than longitudinally tracked samples from a single batch, the observed differences may reflect both storage history and interannual variation in initial tea composition. The common tea garden, cultivar, processing procedure, and storage conditions reduced controllable variation but could not eliminate this source of confounding. The present study focused on catechin–gallic acid-related changes identified in the untargeted dataset, and employed UHPLC-QqQ-MS/MS targeted quantification to evaluate the variation of these quality-relevant non-volatile compounds across consecutive production year samples.

3.2. Catechin and gallate-related annotated features screened from untargeted dataset

Following the global metabolic analysis shown in Fig. 1, catechin–gallic acid-related annotated features were further manually selected from the full UHPLC-QTOF-MS annotation list (Table S3) according to their compound names and chemical relevance. This screening step narrowed the overall untargeted metabolic profile established in Fig. 1 to a chemically interpretable subset, rather than to constitute an independent untargeted analysis. Accordingly, annotated features chemically assignable to the catechin–gallic acid-related polyphenol system were retained for subsequent interpretation, and their associations with production year were characterized using Spearman correlation. A total of 17 such features were retained, encompassing catechin monomers, catechin gallates, methylated gallates, trigallic acid, theaflavin gallates, and prodelphinidin gallates, and were ranked by the absolute Spearman correlation coefficient with production year (Fig. 2; full annotation and ranking details are provided in Table S3). These features were distributed across a broad range of ranked feature indices and showed both positive and negative Spearman correlation coefficients, indicating that catechin–gallic acid-related constituents did not change unidirectionally across production years but exhibited variation in both direction and magnitude. Among the features shown, (+)-gallocatechin, epigallocatechin gallate, epigallocatechin, gallic acid, L-epicatechin, gallocatechin gallate, and (−)-epicatechin gallate corresponded directly to the catechins and gallic acid targeted in the subsequent quantitative analysis; the remaining features may represent more complex derivatives, oxidation products, or polymerization-related constituents of the catechin–gallic acid system. Previous studies have reported that catechins, phenolic acids, and galloylated derivatives can undergo hydrolysis, oxidation, and rearrangement during tea storage, accompanied by changes in gallic acid and degalloylated products (Huang et al., 2021; Peng & Shahidi, 2023; Qin et al., 2012). The 17 features comprised basic catechin monomers, galloylated catechin derivatives, methylated gallates, theaflavin gallates, and prodelphinidin gallates, showing that the untargeted annotations covered several catechin-related and gallate-related compound classes rather than a single structural group. Seven of these annotations corresponded directly to compounds included in the targeted analysis. C and CG were additionally included to complete four structurally matched epimeric pairs, resulting in a targeted panel of four non-galloylated catechins, four galloylated catechins, and GA. The remaining annotations were retained to describe the broader catechin- and gallate-related chemical context and were not included in the targeted quantification.

Fig. 2.

Catechin-gallic acid-related annotated features from UHPLC-QTOF-MS. Seventeen manually curated catechin–gallic acid-related features are shown according to their absolute Spearman correlation coefficients with production year. Blue points indicate features corresponding to compounds included in the targeted UHPLC-QqQ-MS/MS panel, and red points indicate related derivatives or polyphenolic constituents outside the targeted panel. Annotation details are provided in Table S3. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.3. Targeted quantification of catechins and gallic acid

To further verify the catechin–gallic acid-related changes suggested by the untargeted metabolomic screening, eight representative catechins and gallic acid were quantified by UHPLC-QqQ-MS/MS (Fig. 3). The targeted panel included galloylated catechins, namely EGCG, GCG, ECG, and CG; non-galloylated catechins, namely C, EC, GC, and EGC; and gallic acid (GA). All concentrations were expressed on a dry-weight basis. The calibration range and linearity, LOD, LOQ, intra- and inter-day precision, and recovery are summarized in Table 2. The nine target compounds showed good calibration linearity (R² > 0.999).

Fig. 3.

Targeted quantification of catechins and gallic acid across production years. (A) Galloylated catechins. (B) Non-galloylated catechins. (C) Gallic acid. (D) Relative changes in compound groups using RG25 as the reference. Values are expressed as mean ± SD (n = 3); concentrations in A-C are expressed on a dry-weight basis.

Table 2.

Method validation parameters for UHPLC-QqQ-MS/MS quantification of catechins and gallic acid.

Compound	Calibration curve	R2	Linear range (μg/mL)	LOD (μg/mL)	LOQ (μg/mL)	Precision (RSD, %,)		Recovery (%)
						Intra-day (n = 5)	Inter-Day (n = 15)	Intra-day recovery (%)	Inter-day recovery (%)
C	y = 3.63 × 105 x + 9.65 × 103	0.9999	0.1–10	0.019	0.064	2.683	4.709	103.544	98.695
EC	y = 2.87 × 105 x + 1.09 × 104	0.9999	0.1–10	0.007	0.022	1.970	5.829	99.048	95.591
GC	y = 2.46 × 105 x + 1.32 × 104	0.9999	0.1–10	0.013	0.043	0.723	8.215	107.698	102.483
EGC	y = 8.86 × 104 x + 3.67 × 103	0.9999	0.1–10	0.030	0.100	3.056	7.961	95.056	101.479
CG	y = 1.95 × 105 x + 8.57 × 103	0.9999	0.1–10	0.013	0.042	2.692	2.896	96.972	97.227
ECG	y = 1.77 × 105 x + 7.94 × 103	0.9999	0.1–10	0.013	0.042	2.854	4.688	97.466	95.928
GCG	y = 2.23 × 105 x + 9.42 × 103	0.9999	0.1–10	0.004	0.014	4.569	9.986	114.526	106.430
EGCG	y = 1.91 × 105 x + 7.21 × 103	0.9999	0.1–10	0.017	0.057	1.962	7.333	103.624	105.169
Gallic Acid	y = 1.14 × 105 x + 3.84 × 103	0.9999	0.1–10	0.047	0.156	2.609	8.068	99.610	104.459

As shown in Fig. 3A, EGCG and GCG were the dominant galloylated catechins and showed similar variation patterns across production years, with relatively high levels in RG23 and RG18 and lower levels in RG21–RG20 and RG16. ECG and CG were present at lower concentrations but also increased in RG18, indicating partially synchronized fluctuations within the galloylated catechin group. However, these compounds did not show a unidirectional increase or decrease with storage duration. For non-galloylated catechins, EGC was the dominant compound and showed a larger fluctuation amplitude than GC, EC, and C (Fig. 3B). EGC was relatively high in RG24, RG22, and RG18, but lower in RG20 and RG16, whereas GC, EC, and C remained at lower levels with smaller fluctuations and were not fully synchronized with EGC. GA showed a different pattern from catechins: it was lowest in RG25 and increased gradually from RG23 to RG18 (Fig. 3C). When RG25 was used as the reference, GA showed the most pronounced relative increase, rising by 104.45% in RG20 and by 168.25% and 166.25% in RG17 and RG16, respectively (Fig. 3D). These results indicate that the targeted polyphenols in Rougui did not increase or decrease synchronously during long-term storage; instead, galloylated catechins showed relatively coordinated fluctuations, non-galloylated catechins showed compound-specific variation, and GA increased markedly. Further analysis showed that both galloylated (G) and non-galloylated (NG) catechin pools varied non-monotonically across storage durations (Table 3). The G/NG ratio ranged from 3.08 to 5.63, while galloylated catechins accounted for 75.51–84.90% of total catechins, reflecting a redistribution in catechin composition across storage durations.

Table 3.

Quantitative composition of galloylated and non-galloylated catechins across storage durations.

Sample	Storage duration(years)	G (mg/g)	NG (mg/g)	Total catechins (mg/g)	G/NG	Galloylated fraction (%)
RG25	0	77.81 ± 2.77	19.92 ± 0.62	97.73 ± 3.27	3.91 ± 0.09	79.61 ± 0.37
RG24	1	74.86 ± 9.75	24.28 ± 3.07	99.14 ± 12.83	3.08 ± 0.02	75.51 ± 0.11
RG23	2	96.55 ± 15.80	17.14 ± 2.56	113.69 ± 18.35	5.63 ± 0.11	84.90 ± 0.25
RG22	3	90.71 ± 15.90	23.02 ± 3.53	113.73 ± 19.43	3.93 ± 0.10	79.72 ± 0.41
RG21	4	67.13 ± 6.65	16.38 ± 2.06	83.50 ± 8.70	4.11 ± 0.11	80.42 ± 0.43
RG20	5	63.46 ± 1.03	15.22 ± 0.29	78.67 ± 1.32	4.17 ± 0.01	80.66 ± 0.04
RG19	6	81.33 ± 18.09	19.41 ± 3.77	100.73 ± 21.86	4.18 ± 0.12	80.68 ± 0.45
RG18	7	100.74 ± 15.27	24.32 ± 3.27	125.05 ± 18.52	4.14 ± 0.11	80.52 ± 0.41
RG17	8	73.75 ± 11.65	16.80 ± 2.65	90.55 ± 14.30	4.39 ± 0.01	81.45 ± 0.02
RG16	9	65.42 ± 11.10	13.74 ± 2.32	79.15 ± 13.42	4.76 ± 0.01	82.64 ± 0.03

Values are presented as mean ± SD (n = 3). Galloylated catechins (G) = EGCG + GCG + ECG + CG; Non-galloylated catechins (NG) = EGC + GC + EC + C; Total catechins = G + NG; Galloylated fraction = G/(G + NG) × 100%.

The enrichment of GA was consistent with previous findings in aged oolong tea, which showed that marked GA accumulation was associated with changes in gallated catechins and flavonol-related compounds (Lee et al., 2008). Similarly, studies on aged Tieguanyin oolong tea showed that baking and subsequent aging substantially changed the phenolic composition, with baking exerting a particularly strong influence on phenolic changes before long-term storage (Chen et al., 2013). From a broader transformation perspective, galloylated catechins may release GA through hydrolysis or degradation of galloyl groups, while oxidation, isomerization, degradation, and polymerization may also contribute to changes in catechin composition (Abudureheman et al., 2022; Battestin et al., 2008; Peng & Shahidi, 2023; Tanaka & Kouno, 2003). Therefore, the increase in GA observed here is more appropriately interpreted as part of broader catechin–GA-related compositional changes, rather than as the direct linear conversion of a single catechin precursor into GA. From the perspective of taste-related constituents, GA enrichment may be associated with changes in bitterness- and astringency-related non-volatile compounds during Rougui storage. Previous studies have shown that GA and other phenolic acids are associated with bitterness and astringency perception in tea infusions and may interact with catechins to influence oral sensations (Chen et al., 2022; Ren et al., 2022; Ye et al., 2022).

3.4. Correlation matrix of targeted catechins and gallic acid

Although Fig. 3 and Table 3 provide quantitative data, these results alone do not reveal whether the targeted catechin and gallic acid constituents varied in concert as a system across production years. To address this, a Pearson correlation matrix was constructed from sample-level quantitative data for all nine targeted compounds (Fig. 4). Thirty of the 36 pairwise correlations remained significant after Benjamini-Hochberg correction (adjusted P < 0.05; Table S6).

Fig. 4.

Correlations among targeted catechins and gallic acid across individual samples. Pearson correlation analysis was performed using UHPLC-QqQ-MS/MS (n = 30). The upper triangle shows Pearson correlation coefficients, the lower triangle shows pairwise scatter plots with fitted regression lines, and the diagonal panels show compound distributions. Asterisks indicate significance based on Benjamini-Hochberg-adjusted P values across all 36 unique pairwise comparisons: *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

The results revealed structured inter-compound associations within the catechin–gallic acid system, rather than independent variation of individual compounds. EGCG–GCG and ECG–CG showed near-perfect positive correlations (r = 0.9995 and 0.9987, respectively), and C–GC showed a strong positive correlation (r = 0.9516), while EC–EGC, EGC–EGCG, and EGC–GCG showed moderately strong positive correlations (r = 0.8115, 0.7516, and 0.7505, respectively). These findings indicate that structurally related catechins varied in a coordinated manner across consecutive production-year samples, consistent with the compositional redistribution of the catechin pool described in Section 3.3.

The near-perfect correlations for EGCG–GCG (r = 0.9995) and ECG–CG (r = 0.9987) may reflect their C-2 epimeric relationships, which could confer similar susceptibility to hydrolysis or oxidative transformation during storage (Peng & Shahidi, 2023; Sang et al., 2005; Suzuki et al., 2003). Sample-level verification showed that data points for both pairs did not coincide with the y = x reference line and that the fitted slopes were both below 1 (Fig. S1–S2), confirming that the paired compounds are quantitatively distinguishable and do not represent analytical duplicates or signal misassignments. These correlations are therefore interpreted as evidence of coupled concentration variation across production years, rather than proof of direct chemical interconversion.

By contrast, GA showed near-zero linear correlation with both EGCG and GCG (r = 0.0026 and 0.0037, respectively; Fig. 4). This result does not exclude the potential contribution of galloylated catechin hydrolysis or degradation to GA formation (Lee et al., 2008; Li et al., 2013), but indicates that the storage-related variation in GA cannot be attributed to a linear response to any single galloylated catechin precursor. Considered alongside the relative changes in GA compared with RG25 shown in Fig. 3D, the near-zero correlation suggests that GA accumulation may reflect the combined influence of release from multiple galloylated substrates, catechin derivative transformations, and subsequent oxidation, polymerization, or consumption processes, rather than a single dominant pathway. Consistently, increases in simple phenolic acids accompanied by decreases in phenolic acid esters have been reported during Pu-erh tea fermentation (Ge et al., 2019). However, the present data do not establish a specific GA formation pathway. Beyond the compounds subjected to targeted quantification, the untargeted annotated features in Fig. 2 included methylated gallates, galloylated catechin derivatives, theaflavin gallates, and prodelphinidin gallates — constituents not covered by the targeted panel. These annotations extend the chemical scope beyond the targeted panel. Previous studies have shown that catechin oxidative coupling can yield theaflavins, theasinensins, dimeric catechin oxidation products, and other secondary polyphenols (Li et al., 2022; Neilson et al., 2010; Tanaka et al., 2009).

In this study, Fig. 2 includes annotations of polyphenolic derivatives beyond the nine compounds quantified in Fig. 3, whereas Fig. 4 shows correlations among the quantified compounds. The results indicate differences among production-year samples in catechin composition and GA content; however, the overall variation did not follow a consistent upward or downward trend, and these observations do not establish precursor–product relationships or specific transformation pathways. These compositional profiles may provide reference data for comparing the chemical consistency of stored Rougui tea products and for selecting measurable targets in future quality-control studies.

4. Conclusion

This study integrated untargeted metabolomics and targeted quantification to characterize storage-associated non-volatile chemical variation in Wuyi rock tea Rougui across ten consecutive production years. The production-year series showed no consistent year-to-year increase or decrease in individual compounds; instead, the batches differed in their overall metabolic composition. Untargeted screening localized these metabolic differences to a catechin-gallic acid-related polyphenol domain encompassing catechins, gallates, and their derivatives. Further targeted quantification revealed compositional redistribution between galloylated and non-galloylated catechin fractions.

GA was higher in several earlier production-year batches. Correlation analysis further indicated coordinated variation among several catechins, while the broader untargeted annotations included additional catechin- and gallate-related derivatives but did not demonstrate linear conversion of individual catechin precursors.

Although inter-annual variability among independently manufactured historical batches cannot be completely excluded, these findings improve current understanding of storage-associated non-volatile compositional variation in aged Wuyi rock tea Rougui and provide a chemical basis for its future quality evaluation.The combined catechin-pool and GA profiles may provide a reference for comparing the chemical consistency of stored Rougui samples.

CRediT authorship contribution statement

Yuansen Qi: Writing – original draft, Visualization, Investigation, Formal analysis, Data curation. Keqin Huang: Methodology, Investigation. Chunhua Ma: Writing – review & editing, Supervision, Funding acquisition. Yan Huang: Supervision, Funding acquisition.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data will be made available on request.