Lipid oxidation driven olefinic aldehyde biosynthesis shapes aged aroma in Qingzhuan tea

Abstract

This study systematically investigates lipid dynamics and their role in aroma formation during Qingzhuan tea (QZT) processing. Using UHPLC-MRM-MS/MS and GC–MS, we analyzed fatty acids (FAs) and oxidized fatty acids (OFAs) across seven processing stages, identifying 31 FAs and 55 OFAs. Polyunsaturated fatty acids (PUFAs), particularly α-linolenic acid (C18:3) and linoleic acid (C18:2), dominated the lipid profiles (43.7 %–60.1 %), exhibiting biphasic dynamics: a 5.3-fold increase during pile fermentation and natural aging (RT → A12) followed by oxidative degradation (30.0 % reduction in QZT). Multivariate analysis revealed 76 differential lipids correlating with 22 key volatiles, including (E,E)-2,4-heptadienal and (E)-2-octenal. Metabolic pathway analysis mapped lipoxygenase/cyclooxygenase (LOX/COX)-mediated oxidation of C18:3/C18:2 to hydroperoxides, which were then cleaved by lyases into aldehydes. Isotope labeling confirmed cross-pathway interactions between linoleic and arachidonic acid metabolism, while modeling experiments validated enzymatic generation of C6–C9 aldehydes from lipid precursors. This work elucidates the biochemical basis of QZT's aged aroma, providing actionable insights for flavor modulation in fermented teas.

Highlights

• •Processing shapes Qingzhuan tea's fatty acid profile, dominated by polyunsaturated fatty acids.
• •Oxidized fatty acids increase during processing, especially in pile fermentation.
• •Unsaturated aldehydes and fatty acid variations are closely linked.
• •Cross-pathway interactions confirmed via isotope tracing.
• •PUFA oxidation drives Qingzhuan tea's characteristic aged aroma.

1. Introduction

Qingzhuan tea (QZT), a geographically protected dark tea originating from Hubei Province, China, holds both cultural and economic significance (Liu et al., 2022). Its characteristic flavor profile and health-promoting properties are attributed to a complex matrix of bioactive compounds, including polyphenols, flavonoids, and tea pigments (Feng et al., 2020; Fu et al., 2022). The formation of its characteristic aroma involves complex biochemical transformations during processing, particularly through lipid oxidative degradation, glycoside hydrolysis, and Maillard reactions (Wang et al., 2022).

Notably, lipids (4 %–9 % dry weight), particularly unsaturated fatty acids such as linolenic and linoleic acids, serve as critical precursors for aroma generation (Schuh & Schieberle, 2006; Zhou et al., 2022). Mechanistic studies across tea varieties have established that lipid hydrolysis releases free fatty acids, which subsequently undergo enzymatic or non-enzymatic oxidation to form aroma-active aldehydes and ketones (Wang et al., 2021). Cross-tea comparative studies reveal distinct lipid-aroma relationships: In green tea, linoleic acid cleavage generates (Z)-3-hexenal, the progenitor of C6 aldehydes/alcohols (Kim, Ryu, Roh, Lee, & Park, 2003). Black tea processing elevates (Z)-3-hexenol and (E)-2-hexenal through glyceride remodeling (Li et al., 2017). Oolong tea biosynthesis produces methyl jasmonate, (Z)-jasmonate, and jasmine lactone via LOX-catalyzed cyclization (Ho, Zheng, & Li, 2015).

These lipid-derived aromatics are modulated by both intrinsic factors (cultivar genetics, leaf maturity) (Ravichandran & Parthiban, 2000) and processing parameters—particularly during critical stages like fixation, rolling, and fermentation—where thermal and mechanical stresses accelerate lipid peroxidation (Guo, Chen, Guo, & Lin, 2022; Li et al., 2017; Li et al., 2021). Hydroperoxide intermediates formed during oxidation subsequently fragment into key odorants, including furans and α,β-unsaturated aldehydes (Huang et al., 2023), establishing lipid metabolism as a central hub for tea aroma engineering.

Despite these advances, the lipid metabolic network in QZT's unique pile fermentation system remains poorly characterized. Our previous work identified a class of olefinic aldehydes—including (E,E)-2,4-heptadienal, (E,E)-2,4-nonadienal, and (E)-2-decenenal—as signature aroma contributors that significantly accumulate during pile fermentation (Liu et al., 2022). While lipoxygenase (LOX)-mediated enzymatic pathways and non-enzymatic autoxidation are hypothesized to drive their formation (Feng et al., 2019; Ho et al., 2015), three fundamental knowledge gaps hinder mechanistic understanding: (1) the dynamic evolution of lipid species (e.g., oxidized derivatives) during QZT processing stages, (2) the specific precursor-product relationships governing olefinic aldehyde biosynthesis, and (3) the regulatory effects of processing parameters on lipid-aroma interactions.

To resolve these limitations, we implemented an integrated lipidomics strategy combining gas chromatography–mass spectrometry (GC–MS) and ultra-high performance liquid chromatography coupled with tandem mass spectrometry (UHPLC-MS/MS). This approach enabled systematic mapping of lipid profiles across manufacturing stages, identification of key precursors through temporal correlation analysis, and experimental validation of metabolic pathways via simulated reactions. Our findings reveal previously unrecognized lipid transformation mechanisms in dark tea fermentation, providing a theoretical framework for aroma chemistry in fermented teas. Furthermore, the established lipid-aroma correlation model offers practical insights for optimizing QZT processing techniques through targeted lipid metabolism regulation.

2. Materials and methods

2.1. Chemicals

Chemical reagents were obtained from certified suppliers as follows: Analytically pure isopropanol and methanol from Sinopharm Chemical Reagent (Shanghai, China); a 49-component fatty acid methyl ester (FAME) blend from ANPEL Scientific Instrument (Shanghai, China); chromatographic standards including N-hexane, α-linolenic acid, linoleic acid, and deuterated stearic acid (d35-stearic acid, ≥98 %) from Sigma-Aldrich (St. Louis, MO, USA); ethyl decanoate (internal standard) and N-alkanes (C5-C30) from Sigma-Aldrich (Shanghai, China); trimethylsilyldiazomethane in hexane (2.0 M) from Macklin Biochemical (Shanghai, China); ¹³C-labeled α-linolenic acid (98 %) from Cambridge Isotope Laboratories (USA); and a manual solid-phase microextraction (SPME) assembly equipped with Carboxen/divinylbenzene/polydimethylsiloxane fiber (CAR/DVB/PDMS, 50/30 μm, 2 cm) from Supelco (Bellefonte, PA, USA).

2.2. Sample preparation

Tea leaves from Zhaoliqiao Tea Factory (Chibi City, Hubei Province) were collected during seven manufacturing stages: raw tea (RT), first turning (FT), second turning (ST), third turning (TT), natural 6-month aging (A6), natural 12-month aging (A12), and final dried product (QZT). Processing strictly adhered to the protocol established by Liu et al. (2022), encompassing four sequential phases: Fresh tea shoots first underwent enzyme inactivation via rotary fixation at 250 °C for 3 min, followed by rolling (10 min) and sun-drying (2 days) to obtain RT. The RT material was then moistened to 30 %–32 % water content and subjected to pile fermentation (5.0 × 10.0 × 3.0 m dimension) with triplicate turning at 10-day intervals (FT/ST/TT stages). Subsequently, fermented leaves entered natural aging at ambient temperature for 6–12 months. Finally, aged tea was refined, then steamed, pressed into bricks, and dried through a graduated temperature protocol (35 °C → 65 °C over 7–10 days) to yield QZT. Samples underwent vacuum freeze-drying prior to lipidomic and volatile characterization.

2.3. Determination of free fatty acids using GC–MS

Free fatty acids (FAs) were quantified by GC–MS following a modified protocol (Dao, Zhang, Wang, & Wang, 2023). Samples (100 mg) were homogenized in 460 μL isopropanol/hexane (2:3, v/v) with 40 μL internal standard (ethyl decanoate, 1 mg/mL in hexane), vortexed (10 s), and ultrasonicated (5 min, ice bath). After centrifugation (12,000 ×g, 4 °C, 15 min), supernatants were collected and re-extracted with 500 μL extraction solvent. Combined supernatants (400 μL) were concentrated under N₂, derivatized with 200 μL methanol and 100 μL trimethylsilyldiazomethane (room temperature, 15 min), then reconstituted in 160 μL hexane for GC–MS analysis.

$$\text{The calculation formula is}\mathrm{as}\text{follows}:{\mathrm{C}}_{\mathrm{i}}=\left({\mathrm{C}}_1\times {\mathrm{V}}_1\times {\mathrm{V}}_2\right)/\left(\mathrm{M}\times {\mathrm{V}}_3\right)$$ (1)

In eq. (1), C_i: The content of FAs in the sample (μg/g); C₁: Concentration of FAs in the secondary solution (mg/L); V₁: Dissolution volume (mL); V₂: Volume of added extraction solution (mL); V₃: Volume of extracted solution taken out; M: Weighing (mg).

GC–MS analysis was performed on an Agilent 7890B/5977B system equipped with a DB-FastFAME column. Operational parameters included: 1 μL injection volume at 5:1 split ratio; injector and detector temperatures maintained at 240 °C and 230 °C respectively; and a multi-ramp oven program—50 °C (1 min) → 200 °C at 50 °C/min (15 min) → 210 °C at 2 °C/min (1 min) → 230 °C at 10 °C/min (16.5 min). Mass spectra were acquired in Scan/SIM mode (m/z 30–400) following a 7-min solvent delay.

2.4. Determination of oxidized fatty acids using UHPLC-MS/MS

Oxidized fatty acids (OFAs) were quantified following a modified Lv and Liu (2023) protocol. Briefly, 50 mg tea samples underwent methanol extraction with ultrasonication and centrifugation. The supernatants were purified by solid-phase extraction: cartridge-loaded, washed with 1 mL 5 % (v/v) methanol/water, nitrogen-evaporated, derivatized at 100 °C, and reconstituted in 30 % (v/v) acetonitrile/water. After final centrifugation, extracts were analyzed via UHPLC-MS/MS (EXIONLC System; Sciex) fitted with an ACQUITY UPLC BEH C18 column (150 × 2.1 mm, 1.7 μm; Waters). Chromatographic conditions employed 0.01 % aqueous formic acid (A) and acetonitrile (B) as mobile phases, with column temperature at 50 °C, autosampler at 4 °C, and 10 μL injection volume. Multiple reaction monitoring (MRM) parameters were optimized per flow injection analysis (Yuan et al., 2022).

2.5. Modeling experiments analysis of target lipid precursors

We selected α-linoleic acid and linoleic acid as the target lipid precursors, which were prepared as a 100 mg/mL solution. The reaction conditions were designed to simulate the temperature parameters of traditional QZT pile fermentation (the temperature increases from 30 °C to 60 °C over a span of 2 days, then stabilizes at 60 °C for the subsequent 15 days). Additionally, we separately added 2 mg of lipoxygenase (50,000 u/mg) to investigate the impact of enzymatic action on fatty acid degradation. All other reaction conditions remained unchanged. Volatile components were analyzed via HS-SPME/GC–MS post-simulation reaction.

2.6. Qualitative and quantitative analysis of the volatiles by GC–MS analysis

Volatile compounds were extracted by headspace-SPME and analyzed via GC–MS per Liu et al. (2022). Briefly, 3 g tea samples infused with 150 mL boiling water in 500 mL vials were spiked with ethyl decanoate (20 μL, 71 μg/g), equilibrated at 60 °C (5 min), then exposed to CAR/DVB/PDMS fiber for 50 min at identical temperature. Analytes were thermally desorbed in splitless mode (240 °C, 3.5 min) using an HP-5MS column (30 m × 0.25 mm, 0.25 μm; J&W Scientific) with helium carrier gas (99.999 % purity, 1 mL/min constant flow). The temperature program initiated at 50 °C (5 min), ramped to 180 °C at 3 °C/min (2 min hold), then to 250 °C at 10 °C/min (3 min hold). Mass spectra were acquired from m/z 35–450 AMU.

2.7. Analysis of ¹³C-labeled α-linolenic acid derived metabolites by UHPLC-MS/MS

One kilogram of raw tea was spiked with [¹³C]α-linolenic acid to achieve 30 % moisture content and subjected to staged pile fermentation under the conditions detailed in Section 2.5, with samples collected at both mid-fermentation and the final stage. Metabolite extraction entailed homogenizing 20 mg aliquots in ice-cold methanol/water (3:1, v/v) containing isotopic standards through three cycles of mechanical disruption (35 Hz, 4 min) and ice-bath sonication (5 min), followed by protein precipitation (−40 °C, 1 h) and centrifugation (13,800 ×g, 15 min, 4 °C). Supernatants were analyzed by UHPLC-MS/MS (Vanquish-Orbitrap Exploris 120 system) using a HSS T3 column (2.1 × 100 mm, 1.8 μm) with mobile phase A [5 mmol/L ammonium acetate/acetic acid (aq)] and B (acetonitrile). Mass spectrometry operated in IDA mode (Xcalibur) under key parameters: ESI source (±3.8/−3.4 kV; sheath/aux gas 50/15 Arb; capillary 350 °C), resolutions 60,000 (MS) and 15,000 (MS/MS), and stepped normalized collision energies (10/30/60 eV). Peak quantitation and mass isotope distribution analysis employed AccuCor (v0.2.4).

2.8. Statistical analysis

The mean values from three replicates were reported, and SPSS (version 19.0, SPSS Inc., Chicago, IL) was employed for one-way ANOVA to assess significant differences. Metabolites were analyzed using SIMCA (V14.1, Umetrics AB, Umea, Sweden) and annotated via http://www.kegg.jp/kegg/compound/, followed by mapping to the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database. Heatmaps and bar charts were constructed with Origin 2021 (Origin-Lab, Northampton, MA).

3. Results and discussion

3.1. Variations in the fatty acid content during QZT processing

A total of 31 fatty acids were identified across seven processing stages (RT, FT, ST, TT, A6, A12, QZT), comprising 11 saturated (SFAs), 12 monounsaturated (MUSFAs), and 8 polyunsaturated fatty acids (PUFAs) (Table S1-S2). Total fatty acid content exhibited biphasic dynamics: a 4.2-fold surge from 709.31 μg/g (RT) to 2999.23 μg/g (A12), followed by a 27 % reduction in the final product (QZT, 2188.92 μg/g). Among the detected fatty acids, (Z,Z,Z)-9,12,15-linolenic acid (α-linolenic acid, C18:3) had the highest content (196.81–1156.39 μg/g), followed by palmitic acid (C16:0, 205.68–792.64 μg/g) and (Z,Z)-9,12-linoleic acid (linoleic acid, C18:2, 95.81–473.86 μg/g), accounting for 70.2 %–81.9 % of the total profiles in all samples. This tripartite dominance aligns with lipid profiles reported in oolong and green teas (Chen et al., 2022; Zhou et al., 2022), confirming evolutionary conservation of core lipid metabolism across tea varieties.

PUFAs demonstrated the most pronounced processing-induced variation, increasing 5.3-fold from RT to A12, surpassing MUSFAs (3.3-fold) and SFAs (3.4-fold). However, during final processing (A12 → QZT), PUFAs underwent marked degradation (28.9 % reduction), contrasting with minor declines in MUSFAs/SFAs (<5 %). The unsaturated-to-saturated fatty acid ratio (U/S) peaked at TT, reflecting maximal oxidative activity during pile fermentation, and minimized at RT. The contents of MUSFAs, SFAs, and PUFAs showed a significant increase during pile fermentation and natural aging, followed by a decrease in the final stage (Table S1). Notably, this trajectory diverges from green tea and oolong tea processing (Guo et al., 2022), where PUFAs decline continuously while SFAs accumulate (Zhou et al., 2022), highlighting technique-specific lipid remodeling.

High-unsaturation PUFAs exhibited preferential degradation during late-stage processing (A12 → QZT): α-linolenic acid (C18:3) and linoleic acid (C18:2) showed maximum conversion rates (34.8 % and 21.9 % reductions, respectively), consistent with their roles as volatile aldehyde precursors (Wan & Xia, 2015). Paradoxically, select long-chain species—lignoceric acid (C24:0, +25.0 %), (all-Z)-5,8,11,17,17-eicosapentaenoic acid (+617.5 %), and (E)-11-eicosenoic acid (+55.6 %)—displayed anomalous accumulation. This divergence may principally arises from the interconnected mechanisms: Saturated very-long-chain fatty acids (e.g., C24:0) resist thermal degradation due to inherent physicochemical properties—notably higher melting points and enhanced oxidative stability—that minimize volatilization during thermal processing (Guo et al., 2022); concurrently, microbial enzymatic systems drive fatty acid remodeling through malonyl-CoA-dependent elongation of C16-C18 precursors, facilitating neosynthesis of ultra-long-chain species such as C24:0 (Wenning et al., 2019). Critically, dynamic modeling revealed stage-dependent regulatory shifts: hydrolysis-dominated esterase activity prevailed during pile fermentation (RT → A12), while lipoxygenase (LOX)-mediated β-oxidation became dominant post-aging (A12 → QZT), collectively driving the observed lipid trajectory.

The observed lipid dynamics arose from competing hydrolysis-oxidation equilibria. Initial stages (RT → A12) featured rapid hydrolysis of bound lipids into free fatty acids, outpacing oxidative losses. Subsequent aging (A12 → QZT) activated LOX pathways, preferentially oxidizing PUFAs into aldehydes/ketones while sparing SFAs due to their oxidative inertness. MUSFAs exhibited intermediate susceptibility, with degradation rates 40 % lower than PUFAs. The 18 % decline in the U/S ratio during aging underscores PUFAs' enzymatic vulnerability, likely mediated by microbial lipases during dark fermentation. These phase-specific transformations illuminate how post-fermentation biochemistry tailors lipid-derived aroma precursors in QZT.

3.2. Variations in the oxidized fatty acid content during QZT processing

Employing a UHPLC-MRM-MS/MS platform, we systematically mapped 55 OFAs across six subclasses: α-linolenic acid derivatives (ALAs), linoleic acid derivatives (LAs), arachidonic acid derivatives (ARAs), eicosapentaenoic acid derivatives (EPAs), docosahexaenoic acid derivatives (DHAs), and other minor species (Table S1, Table S3). ALAs and LAs dominated the OFA pool, collectively constituting 92.65 %–98.59 % of total oxidized lipids (ALAs: 47.27 %–62.22 %; LAs: 35.90 %–45.56 %).

Total OFA content exhibited stage-dependent escalation, culminating in a 3.9-fold surge during aging (A6 → A12). Early aging phases (A6) triggered pronounced accumulation of ALAs, LAs, and EPAs, while ARA levels peaked specifically during pile fermentation (FT-TT). Conversely, DHA derivatives showed no statistically significant fluctuations (p > 0.05) across processing stages (Fig. S1).

The observed OFA patterns align with established lipoxygenase (LOX) and cyclooxygenase (COX) catalytic mechanisms, where PUFAs generate >260 oxidative metabolites including hydroxyls, peroxides, ketones, and epoxides (Astarita, Kendall, Dennis, & Nicolaou, 2015; Milic, Hoffmann, & Fedorova, 2013). The progressive OFA accumulation—particularly during aging—suggests sustained enzymatic activity (LOX/COX) and non-enzymatic autoxidation of liberated PUFAs (Fu, Qu, Yang, & Zhang, 2016; Mosblech, Feussner, & Heilmann, 2009). Notably, ALA and LA emerge as critical flavor precursors, their high oxidation susceptibility driving the formation of characteristic volatiles through LOX/COX pathways. Mechanistically, ALA oxidation generates (E,E)-2,4-heptadienal derivatives, while LA degradation yields (E)-2-octenal compounds, collectively constituting the molecular foundation of QZT's aged aroma profile.

3.3. Screening for the key lipid compounds during QZT processing

Multivariate statistical analysis was employed to identify critical differential lipid components (FAs and OFAs) across QZT processing stages. The principal component analysis (PCA) model (Fig. 1A) revealed that principal components 1 and 2 explained 51.87 % and 18.91 % of total variance, respectively, effectively distinguishing four distinct processing phases: RT (Stage I), FT-ST-TT cluster (Stage II), A6-A12 group (Stage III), and QZT (Stage IV).

Fig. 1.

Score scatter plot for the PCA model (A), upset plot (B) of differential compound among groups FT vs. RT, ST vs. FT, TT vs. ST, A6 vs. TT, A12 vs. A6, QZT vs. A12, and volcano plot (C1-C6). The differential compounds were screened out based on the criteria of variable importance in projection (VIP) > 1, p < 0.05, and |log₂FC| > 0.26. In upset plot, the red bar chart represents the total number of different substances in groups; the blue bar chart represents the number of different substances shared in groups. In volcano plots, the red dots represent upregulated differential substances, green dots represent downregulated differential substances, and gray dots represent non-significant substances. RT: raw tea, FT: first turning, ST: second turning, TT: third turning, A6: natural 6-month aging, A12: natural 12-month aging, QZT: final dried product of Qingzhuan tea. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

We analyzed the significant distinctions between processes to understand their impact on FAs and OFAs at various stages of QZT processing. To compare FAs and OFAs profiles among samples during processing, we utilized orthogonal partial least squares-discriminant analysis  (OPLS-DA). The OPLS-DA models showed excellent variance explanation and predictive capability across different processing stages: FT vs. RT, ST vs. FT, TT vs. ST, A6 vs. TT, A12 vs. A6, and QZT vs. A12. By applying selection criteria (VIP > 1, p < 0.05, and |log₂FC| > 0.26), we identified 76 distinct lipid compounds (Table 1), visualized their distribution within FAs and OFAs using a volcano plot, and delineated their differential expression patterns across the stages. Comparison among processing stages revealed varying numbers of differential metabolites: FT vs. RT (41), ST vs. FT (21), TT vs. ST (13), A6 vs. TT (36), A12 vs. A6 (4), and QZT vs. A12 (38), with differential upregulation and downregulation patterns as indicated (Fig. 1C1—C6).

Table 1.

A total of 29 differential FAs and 47 differential OFAs during the manufacturing of Qingzhuan tea.

NO.	Compounds	RT vs. FT			FT vs. ST			ST vs. TT			TT vs. A6			A6 vs. A12			A12 vs. QZT
		VIP	Log2FC	Type	VIP	Log2FC	Type	VIP	Log2FC	Type	VIP	Log2FC	Type	VIP	Log2FC	Type	VIP	Log2FC	Type
F1	Myristic acid	1.25	1.07	up													1.22	−0.90	down
F2	Pentadecanoic acid	1.26	1.03	up							1.19	0.28	up				1.21	−0.44	down
F3	Palmitic acid	1.26	1.33	up							1.23	0.30	up				1.26	−0.50	down
F4	Heptadecanoic acid	1.26	0.95	up	1.37	0.28	up
F5	Stearic acid	1.26	0.73	up	1.37	0.27	up
F6	(E)-Nonadecenoic acid	1.22	0.60	up
F7	Arachidic acid	1.23	0.55	up
F8	Heneicosanoic acid	1.21	0.38	up							1.21	0.35	up
F9	Behenic acid	1.24	0.71	up
F10	Tricosanoic acid	1.24	0.65	up
F11	Lignoceric acid	1.24	0.68	up													1.23	0.32	up
F12	(Z)-9-Myristoleic acid	1.19	−1.58	down
F13	(E)-9-Tetradecenoic acid	1.19	0.82	up													1.15	−0.89	down
F14	(Z)-10-Pentadecenoic acid	1.25	0.78	up							1.22	0.26	up				1.25	−1.47	down
F15	(Z)-9-Palmitoleic acid	1.21	0.44	up													1.26	−0.44	down
F16	(E)-9-Palmitelaidic acid							1.33	1.10	up				1.78	−0.51	down
F17	(Z)-10-Heptadecenoic acid	1.26	1.12	up													1.25	−0.35	down
F18	(Z)-9-Octadecenoic acid	1.25	1.37	up	1.38	0.41	up										1.26	−0.36	down
F19	(Z)-11-Octadecenoic acid	1.25	1.04	up	1.36	0.30	up
F20	(Z)-11-Eicosenoic acid	1.21	0.83	up	1.35	0.32	up
F21	(E)-11-Eicosenoic acid										1.19	0.40	up
F23	(E,E)-9,12-Linolelaidic acid							1.51	−0.42	down	1.19	0.41	up				1.23	−0.27	down
F24	(Z,Z)-9,12-Linoleic acid (Linoleic acid)	1.27	1.78	up	1.38	0.39	up										1.27	−0.36	down
F25	(Z,Z,Z)-6,9,12-Linolenic acid (γ-Linolenic acid)	1.26	−1.42	down													1.27	1.49	up
F26	(Z,Z,Z)-9,12,15-Linolenic acid(α-Linolenic acid)	1.27	2.04	up	1.38	0.41	up										1.28	−0.62	down
F28	(all-Z)-11,14,17-Eicosatrienoic acid	1.26	1.10	up	1.37	0.53	up										1.18	−0.26	down
F29	(all-Z)-5,8,11,14,17-Eicosatetraenoic acid	1.22	−1.48	down
F30	(all-Z)-4,7,10,13,16-Docosapentaenoic acid							1.54	−0.67	down	1.19	0.37	up
F31	(Z)-15-Nervonic acid				1.29	0.41	up										1.24	0.64	up
OF1	5S-HETrE	1.16	1.42	up							1.18	0.82	up
OF3	9S-HOTrE	1.22	0.80	up							1.22	1.86	up
OF4	13S-HOTrE	1.23	0.94	up							1.21	2.29	up				1.22	0.41	up
OF5	13S(γ)-HOTrE										1.18	2.53	up
OF6	15S-HETrE	1.15	1.25	up	1.27	0.39	up				1.19	1.44	up				1.27	1.05	up
OF7	9-OxoODE	1.20	0.53	up							1.14	1.29	up				1.22	0.63	up
OF8	9S-HODE	1.25	0.60	up							1.23	1.98	up				1.22	0.47	up
OF9	±9(10)-EpOME										1.23	1.35	up
OF10	±9(10)-DiHOME				1.37	0.56	up	1.56	0.71	up	1.22	1.00	up				1.27	1.62	up
OF11	±12(13)-EpOME	1.18	0.48	up							1.23	1.21	up
OF12	±12(13)-DiHOME	1.17	1.64	up	1.29	0.63	up	1.46	0.47	up	1.14	0.91	up				1.26	1.10	up
OF13	13-OxoODE							1.52	0.35	up	1.22	2.17	up
OF14	13S-HODE	1.13	0.43	up							1.20	2.33	up				1.19	0.33	up
OF18	12S-HEPE										1.20	10.16	up
OF19	15S-HEPE										1.12	0.69	up
OF20	ARA	1.19	0.87	up
OF21	9R-HETE										1.22	9.95	up	1.73	1.40	up
OF23	12S-HETE										1.17	1.30	up
OF25	19S-HETE	1.25	9.82	up													1.22	1.26	up
OF26	6-(E)-LTB4										1.23	8.04	up				1.19	0.65	up
OF27	LTE4	1.17	−7.38	down
OF28	14,15-LTE4										1.21	11.43	up
OF29	14,15-LTD4				1.29	8.71	up	1.46	−8.71	down	1.19	8.30	up				1.23	1.27	up
OF30	PGB2																1.14	1.15	up
OF31	PGD2				1.19	1.87	up
OF32	13,14-dihydro-15-keto PGD2										1.22	9.40	up
OF33	11β-PGE2				1.38	12.59	up	1.55	0.77	up	1.22	0.54	up	1.68	0.64	up	1.25	−1.75	down
OF34	2,3-dinor-8-isoPGF2α	1.20	10.65	up	1.30	−10.65	down							1.75	12.33	up	1.25	−11.91	down
OF36	13,14-dihydro-15-keto PGF2α										1.23	2.65	up				1.26	0.72	up
OF37	13,14-dihydro PGF2α				1.31	1.06	up	1.54	−13.61	down
OF38	8-iso-15-keto PGF2β																1.11	1.81	up
OF39	20-COOH-AA										1.21	10.50	up
OF40	2,3-dinor TXB2										1.15	13.55	up				1.15	0.87	up
OF41	20-carboxy LTB4	1.05	0.97	up
OF42	5S,6R-DiHETE	1.21	−9.98	down													1.26	−1.50	down
OF43	PGF3α				1.36	0.66	up				1.22	1.43	up				1.22	0.94	up
OF44	DHA				1.36	1.32	up	1.46	−0.91	down
OF45	15-OxoEDE										1.21	1.89	up				1.13	0.37	up
OF46	15-OxoETE	1.21	8.66	up
OF48	9-Nitrooleic acid																1.20	−0.55	down
OF49	DTA							1.33	1.03	up	1.07	−1.72	down
OF50	±5,6-DiHETrE							1.51	8.89	up							1.09	0.59	up
OF51	±11,12-DiHETrE				1.23	2.73	up
OF52	±14,15-DiHETrE										1.20	5.73	up				1.19	0.81	up
OF53	PGE1	1.21	2.24	up
OF54	PGF1α				1.36	9.64	up	1.48	1.12	up
OF55	Δ17–6-keto PGF1α																1.11	−0.34	down

The upset plot analysis (Fig. 1B) revealed overlaps and differences in the levels of various FAs and OFAs during processing. We identified a total of 11, 9, 4, and 2 unique differential metabolites in the FT vs. RT, A6 vs. TT, QZT vs. A12, and ST vs. FT comparisons, respectively. This shared signature predominantly comprised medium-chain saturated FAs (e.g., myristic acid, palmitic acid), very-long-chain saturated FAs (e.g., lignoceric acid), monounsaturated FAs (e.g., palmitoleic acid), key polyunsaturated FAs (e.g., γ-linolenic acid), and critical lipoxygenase (LOX) pathway derivatives—including hydroxides and ketones originating from α-linolenic acid (ALAs: 13S-HOTrE, 9-OxoODE), linoleic acid (LAs: 9S-HODE, 13S-HODE), arachidonic acid (ARAs: 19S-HETE), and eicosapentaenoic acid (EPAs: 5S,6R-DiHETE).

Stage-specific lipid signatures were further delineated through comparative analysis. The FT vs. RT transition exhibited unique enrichment in: (i) saturated very-long-chain fatty acids (C20:0–C24:0), indicative of microbial elongation activity; (ii) monounsaturated fatty acids with double bonds at carbon 9 (e.g., myristoleic acid, C14:1Δ9) and polyunsaturated fatty acids with methylene-interrupted double bonds (e.g., C20:5Δ5,8,11,14,17); and (iii) early-stage oxylipins including epoxy/hydroxy derivatives of α-linolenic acid (C18:3Δ9,12,15), linoleic acid (C18:2Δ9,12), and arachidonate pathways. Conversely, QZT vs. A12 uniquely accumulated terminal oxidation products, predominantly prostaglandins/leukotrienes derived from C20:5Δ5,8,11,14,17 and nitro-fatty acid adducts, reflecting advanced oxidative remodeling during aging. Furthermore, five lipid substances, namely (Z,Z)-9,12-linoleic acid, (Z,Z,Z)-9,12,15-linolenic acid, (Z)-9-octadecenoic acid, (all-Z)-11,14,17-eicosatrienoic acid, and 15S-HETrE, were present in FT vs. RT, ST vs. FT, and QZT vs. A12. This conserved cohort signifies core substrates maintaining enzymatic oxidation cascades throughout processing.

Hierarchical cluster analysis (HCA) visualized differential metabolic variation across QZT manufacturing stages (Fig. 2), with red/blue gradients denoting relative abundance in the normalized heatmap. As shown in Fig. 2A, Class I FAs exhibited high levels in RT or QZT, including (Z)-9-myristoleic acid, (Z,Z,Z)-6,9,12-linolenic acid, (E)-9-palmitelaidic acid, (all-Z)-5,8,11,14,17-eicosatetraenoic acid, and (Z)-15-nervonic acid. Class II FAs displayed high levels in FT and ST, specifically (Z)-9-palmitoleic acid and (all-Z)-4,7,10,13,16-docosapentaenoic acid. In contrast, Class III FAs showed high levels in A6, and A12, including lignoceric acid, behenic acid, tricosanoic acid, etc. These fatty acids are primarily formed during pile fermentation and natural aging processes and likely contribute directly to aroma formation. Previous research has suggested that six‑carbon aldehydes and alcohols are derived from the degradation of (Z,Z,Z)-9,12,15-linolenic acid and (Z,Z)-9,12-linoleic acid (Cañoles, Beaudry, Li, & Howe, 2006; Qin et al., 2014).

Fig. 2.

A total of 29 FAs (A) and 47 OFAs (B) were shown on the heat map, selected by VIP > 1, p < 0.05, and |log₂FC| > 0.26. FAs: fatty acids, OFAs: oxidized fatty acids. RT: raw tea, FT: first turning, ST: second turning, TT: third turning, A6: natural 6-month aging, A12: natural 12-month aging, QZT: final dried product of Qingzhuan tea.

In Fig. 2B, Class I OFAs exhibited the highest levels in QZT, including 5S-HETrE, 9R-HETE, and LTE4. Class II OFAs showed high levels in A6 and A12, specifically ±12(13)-EpOME, 5S,6R-DiHETE, and 15S-HEPE. Lastly, Class III OFAs displayed high levels in FT, ST, and TT, including DHA, 13,14-dihydro PGF2α, and Δ17–6-keto PGF1α.

3.4. Correlation analysis between volatiles and lipids

We conducted correlation analyses to explore the relationship between lipids and volatiles in the QZT manufacturing process, leveraging our prior volatile data (Liu et al., 2022). The Spearman correlation results (Fig. 3A) revealed that (Z,Z,Z)-9,12,15-linolenic acid (C18:3), (Z,Z,Z)-6,9,12-linolenic acid (C18:3), (Z,Z)-9,12-linoleic acid (C18:2), oxygenated fatty acids of α-linolenic acids (ALAs) (5S-HETrE, 9S-HOTrE, 13S-HOTrE, 15S-HETrE), and OFAs of linoleic acids (LAs) (9-OxoODE, 9S-HODE, ±9(10)-DiHOME, ±12(13)-EpOME, 13S-HODE) exhibited significant negative correlations (r < −0.9, p < 0.01) with the concentrations of 12 volatile compounds, including octanal, pentanal, decanal, (Z)-citral, benzyl alcohol, phenylethyl alcohol, methyl salicylate, 2-phenethyl acetate, α-farnesene, and two green leaf volatiles, (Z)-3-hexen-1-ol and 1-penten-3-one. Additionally, the content of C18:3, C18:2, OFAs of ALAs, and OFAs of LAs showed significant positive correlations (r > 0.9, p < 0.01) with the concentrations of (E,E)-2,4-heptadienal, (E,Z)-2,6-nonadienal, safranal, 1-heptanol, 6-methyl-5-hepten-2-one, linalool, camphor, naphthalene, and γ-nonalactone.

Fig. 3.

Correlation analysis between FAs, OFAs and volatile compounds (A) and changes in key lipid components (B) during the manufacturing of Qingzhuan tea. Different lowercase letters in bar chart indicate significant differences between samples (p < 0.05, Duncan's test). (In the correlation analysis heatmap, red represents the positive correlation, and blue represents the negative correlation). FAs: fatty acids, OFAs: oxidized fatty acids. RT: raw tea, FT: first turning, ST: second turning, TT: third turning, A6: natural 6-month aging, A12: natural 12-month aging, QZT: final dried product of Qingzhuan tea. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Among these 22 aromatic compounds, 9 were likely derived from lipid degradation, including octanal, (Z)-3-hexen-1-ol, pentanal, decanal, (E,E)-2,4-heptadienal, (E,Z)-2,6-nonadienal, 1-heptanol, (E)-2-decen-1-ol, and 1-penten-3-one. Previous studies have shown that these lipid-derived aromas primarily originate from C18:3 and C18:2 as their precursors, which undergo oxidation to form ALAs and LAs, the oxidized fatty acids. These findings align with results from Nielsen, Larsen, and Poll (2004), who demonstrated that hexanal, heptanal, pentanol, hexanol, (E)-2-pentenal, (E,E)-2,4-heptadienal, and (E)-2-octenal are produced through the metabolism of linoleic acid and linolenic acid. During the pile fermentation period of QZT, the levels of linoleic acid and α-linolenic acid increased rapidly, while oxygenated fatty acids such as 9S-HOTrE, 13S-HOTrE, 9-OxoODE, 9S-HODE, ±12(13)-EpOME, and 13S-HODE also showed significant increases during the aging period (Fig. 3B). Furthermore, key olefinic aldehydes were predominantly formed during the pile fermentation stage (Liu et al., 2022).

The sequential lipid transformation model is thus validated: (1) hydrolysis of bound lipids releases free C18:2/C18:3 during pile fermentation; (2) oxidative conversion of these PUFAs generates OFAs via LOX/COX pathways; (3) secondary cleavage of OFAs produces volatile aldehydes and alcohols. This cascade elucidates the biochemical foundation of QZT's characteristic flavor profile, bridging lipid dynamics with sensory quality development.

3.5. Analysis of metabolic pathways of key lipid components

During the QZT manufacturing process, significant alterations in FAs and OFAs composition arose from pile fermentation and natural aging. Various FAs and OFAs, such as (Z,Z)-9,12-linoleic acid, (Z,Z,Z)-9,12,15-linolenic acid, (Z)-9-octadecenoic acid, (all-Z)-11,14,17-eicosatrienoic acid, 13S-HOTrE, 9-OxoODE, 9S-HODE, 13S-HODE, 19S-HETE, and 15S-HETrE, exhibited significant alterations. KEGG analysis identified seven lipid-related metabolic pathways (Fig. S2): biosynthesis of unsaturated fatty acids, fatty acid biosynthesis, cutin/suberine/wax biosynthesis, linoleic acid metabolism, fatty acid elongation, α-linolenic acid metabolism, and fatty acid degradation.

During pile fermentation and aging, linoleic acid (C18:2) content increased markedly compared to raw tea, attributed to lipoxygenase-mediated glycerophospholipid degradation. This process generated oxidation products including 13-OxoODE, 9-OxoODE, ±12(13)-EpOME, ±9(10)-EpOME, 9S-HODE, and 13S-HODE (Zhang & Zhang, 2020). Similarly, the content of α-linolenic acid also increased during pile fermentation and aging.

Lipoxygenase (LOX) initiates peroxidation of unsaturated long-chain fatty acids, generating lipid hydroperoxides subsequently cleaved by hydroperoxide lyases (HPLs) into characteristic aldehydes including (E)-2-hexanal, (E,E)-2,6-nonadienal, and (E,E)-2,4-decadienal (Yang, Baldermann, & Watanabe, 2013). Linoleic acid (C18:2) primarily yielded 9/13(S)-hydroperoxyoctadecadienoic acids (HpODEs), with minor 10/12(S)-HpODEs (Li et al., 2021). Some of these fatty acid oxides undergo further oxidation to form 9/13(S)-HODE and 9/13-OXODE, while others degrade to generate compounds such as (E,E)-2,4-decadienal, (E)-2-octenal, (Z)-2-heptaneal, and hexanal (Fig. 4A).

Fig. 4.

Proposed formation pathways of the volatile aldehydes produced from lipids: (A) linoleic acid, (B) [¹³C] linolenic acid, (C) ARA, (LOX: lipoxygenase; HPL: hydroperoxide lyase; ADH: alcohol dehydrogenase; ISO: isomer; ARA: arachidonic acid). FT: first turning, ST: second turning, TT: third turning, A6: natural 6-month aging, A12: natural 12-month aging, QZT: final dried product of Qingzhuan tea.

The oxidation and degradation of α-linolenic acid (C18:3) also release oxylipins, primarily leading to the formation of 9(S)-HpOTrE and 13(S)-HpOTrE. These compounds can be further oxidized to 9(S)-HOTrE and 13(S)-HOTrE, and they can also degrade to produce volatile C9 and C6 aldehydes (Fig. 4B) (Sivankalyani, Maoz, Feygenberg, Maurer, & Alkan, 2017). GC–MS analysis confirmed that C6 aldehydes (e.g., 2-hexanal) and C9 aldehydes (e.g., (E,Z)-2,6-nonadienal, (E,E)-2,4-nonadienal, (E)-2-nonenal) constituted the predominant upregulated oxylipin volatiles during QZT pile fermentation and aging.

3.6. Identification of the volatile precursors from FAs through modeling experiments

To establish connections between fatty acids (α-linolenic acid/linoleic acid) and QZT volatiles, we conducted modeling experiments informed by prior research. GC–MS analysis of four treatment groups revealed dynamic changes in 41 volatiles (Fig. 5), spanning acids, aldehydes (saturated/unsaturated), ketones, alcohols, heterocycles, and esters.

Fig. 5.

A heatmap showing the volatiles produced by each fatty acid. ALA, ALA+LOX, LA and LA + LOX represent the α-linolenic acid, α-linolenic acid and lipoxygenase, linoleic acid, linoleic acid and lipoxygenase, respectively.

In the α-linolenic acid reaction system, a total of 12 volatile components were produced, predominantly acids, aldehydes, and ketones. Notable components with higher content included octanoic acid, (E,E)-2,4-heptadienal, and (E,E)-2,4-hexadienal. The introduction of lipoxygenase to the simulation system resulted in the detection of 18 volatile components, with a significant increase in the content of octanoic acid, (E,E)-2,4-hexadienal, and (E,E)-2,4-hexadienal. Additionally, newly generated components included nonanoic acid, 9-oxo-nonanoic acid methyl ester, 2-propionylfuran, and 2-ethylfuran, among others.

Similarly, the linoleic acid reaction system yielded a total of 19 volatile components, primarily acids, aldehydes, ketones, and alcohols. Key components with higher content comprised octanoic acid, (E)-2-octenal, (Z)-2-heptanal, and hexanal. Upon the addition of lipoxygenase to the simulation system, 24 volatile components were detected, with notable increases in the content of octanoic acid, (E)-2-octenal, and (Z)-2-heptanal. Furthermore, newly generated components included nonanoic acid, 6-dodecanone, nonanal, and 2-undecenenal, among others.

Previous studies have confirmed that unstable fatty acids such as α-linoleic acid and oleic acid are prone to oxidation and produce various aroma components, including aldehydes, ketones, hydrocarbons, alcohols, and acids (Chen, Li, Zhu, Ma, & Rong, 2021; Multari, Marsol-Vall, Heponiemi, Suomela, & Yang, 2019). The degradation products of linoleic acid include heptaldehyde, (E)-2-octenal, and 2-n-hexylfuran (Cossignani, Giua, Simonetti, & Blasi, 2014; Giua, Blasi, Simonetti, & Cossignani, 2013). Similarly, the degradation products of α-linolenic acid consist of acetaldehyde, acetone, propionaldehyde, butyraldehyde, 2-pentenal, 2-hexenal, 2,4-heptadienal, and 2,4-nonadienal (Cao et al., 2014). These identified components partially align with the findings of our study, with variations likely stemming from different reaction conditions. Cao et al. (2020) reported that oleic acid undergoes four self-oxidation pathways, producing 8-hydroperoxide (8-ROOH), 9-hydroperoxide (9-ROOH), 10-hydroperoxide (10-ROOH), and 11-hydroperoxide (11-ROOH), which further degrade to produce nonanal, octanal, and decanal. Additionally, research has suggested that propanal, glutaraldehyde, and hexanal may originate from the decomposition of 15/16 hydroperoxides, 13 hydroperoxides, and 12/13 hydroperoxides, respectively (Frankel, Neff, & Selke, 1981), indicating a close relationship between carbonyl compound formation and the attack positions of hydroperoxides.

Different fatty acids produce distinct isomers of hydroperoxides. For instance, methyl oleate generates four hydroperoxide isomers: 8, 9, 10, and 11-hydroperoxides. The oxidation of methyl linoleate produces two isomers (9/13-hydroperoxide), while oxidized methyl linolenate yields four different isomers (9, 12, 13, and 16-hydroperoxide) (Fatemi & Hammond, 1980). Therefore, variations in carbonyl compounds may arise from different types of hydroperoxides originating from fatty acids with varying structures. Simulation test results on the target lipid precursors demonstrated that α-linoleic acid and linoleic acid can undergo enzymatic or non-enzymatic degradation reactions to produce aldehydes and ketones. As shown in Fig. 4, (Z)-2-hexanal and (Z)-2-octenal can be partially reduced to their respective alcohols by aldehyde reductase, resulting in (Z)-2-hexanol and (Z)-2-octenol, respectively (Al-Dalali, Li, & Xu, 2022). Other researchers have also conducted degradation model experiments on linoleic acid and linolenic acid. According to their findings, linoleic acid led to significantly higher levels of hexanal, heptanal, (E)-2-heptenal, (E)-2-octenal, (E,E)-2,4-decadienal, pentanol, and hexanol. On the other hand, linolenic acid resulted in significantly higher levels of (E)-2-pentenal, (E)-2-hexenal, (E,Z)-2,4-heptadienal, (E,E)-2,4-heptadienal, and butanol. These results align closely with the outcomes of our own model experiments (Nielsen et al., 2004).

3.7. Identification of the volatile precursors from FAs through isotope labeling technique

Metabolic flux analysis of [¹³C]α-linolenic acid was conducted using HPLC-QTOF-MS to elucidate its transformation pathways. The workflow began with the acquisition of [M + H]⁺ or [M-H]⁻ ions from metabolites, followed by precise mass determination (<2 ppm error) and isotopic abundance ratio analysis to infer molecular compositions. Structural characterization was achieved through high-resolution MS/MS spectral interpretation and cross-referencing with established metabolite databases. This integrated approach identified 28 isotopically labeled products, delineating the metabolic pathway of α-linolenic acid into three sequential phases (Fig. 4A-C): (1) LOX-mediated oxidation: Initial conversion to 9/13(S)-hydroperoxyoctadecatrienoic acid (HpOTrE) catalyzed by 9/13-lipoxygenase (LOX); (2) Peroxidase reduction: Subsequent transformation of HpOTrE into 9/13(S)-hydroxyoctadecatrienoic acid (HOTrE); (3) HPL cleavage: Final fragmentation by 9/13-hydroperoxide lyase (HPL) to generate aldehyde derivatives, including (E,E)-2,4-heptadienal.

¹³C isotopic tracing further revealed cross-pathway interactions, demonstrating the participation of linoleic acid in arachidonic acid (ARA) metabolism. The linoleic acid cascade was categorized into three progressive stages: (1) Primary oxidation: Formation of 9/13(S)-hydroperoxyoctadecadienoic acid (HpODE), ±9(10)-epoxyoctadecenoic acid (EpOME), and ± 12(13)-EpOME from linoleic acid; (2) Secondary modification: Conversion of 9/13(S)-HpODE to 9/13(S)-hydroxyoctadecadienoic acid (HODE), coupled with hydration of ±9(10)-EpOME and ± 12(13)-EpOME to yield ±9(10)-dihydroxyoctadecenoic acid (DiHOME) and ± 12(13)-DiHOME, respectively; (3) Tertiary oxidation: Oxidation of 9/13(S)-HODE to 9/13-oxooctadecadienoic acid (OxoODE), ultimately producing aldehyde compounds such as (E,E)-2,4-decadienal. These findings systematically map the enzymatic cascades governing lipid-derived volatile formation in QZT, providing mechanistic insights into the biochemical basis of its characteristic aroma profile.

4. Conclusions

This study systematically delineated lipid dynamics and their mechanistic linkages to unsaturated aldehyde formation during Qingzhuan tea (QZT) processing. Lipid profiling revealed substantial compositional shifts, with polyunsaturated fatty acids (PUFAs) exhibiting the most pronounced variations, alongside progressive accumulation of saturated and monounsaturated fatty acid. Oxidized fatty acids (OFAs) demonstrated sustained elevation, underscoring their role as critical aroma precursors. Multivariate statistical analysis (OPLS-DA, VIP >1) identified 76 differential lipids strongly correlated with 22 key volatiles, including signature compounds like (E,E)-2,4-heptadienal and (E)-2-octenal. Metabolic pathway analysis mapped three core enzymatic cascades: LOX/HPL-mediated oxidation: Conversion of α-linolenic acid (C18:3) to (E,E)-2,4-heptadienal via 9/13(S)-HpOTrE intermediates. Epoxide hydration: Linoleic acid (C18:2) degradation to (E,E)-2,4-decadienal through ±12(13)-EpOME/DiHOME derivatives. Carbonyl generation: Tertiary oxidation of HODE/HOTrE to OxoODE/OxoOTrE aldehydes. Isotope labeling techniques provided critical validation: ¹³C tracing confirmed 28 labeled metabolites, resolving cross-talk between linoleic and arachidonic acid pathways. The metabolic flux model demonstrated: α-Linolenic acid progresses through HpOTrE→HOTrE→(E,E)-2,4-heptadienal. Linoleic acid follows HpODE→HODE→OxoODE→(E,E)-2,4-decadienal. These findings establish a biochemical framework where lipid oxidation (LOX/COX/HPL activity) drives QZT's aged aroma through precursor-product relationships. Consequently, this integrated model provides actionable strategies for modulating aged aroma profiles, such as engineering fermentation microbiota to enrich α-linolenic acid (C18:3)/linoleic acid (C18:2) pools or exogenous application of LOX pathway intermediates (e.g., HpODE) to accelerate enzymatic generation of signature aldehydes including (E,E)-2,4-heptadienal and (E,E)-2,4-decadienal. The integrated approach—encompassing lipid profiling, multivariate correlations, pathway mapping, and isotopic verification—provides actionable targets for flavor modulation, advancing both fundamental understanding of fermented tea chemistry and precision control strategies for industrial production.

CRediT authorship contribution statement

Pengcheng Zheng: Writing – review & editing, Writing – original draft, Investigation, Conceptualization. Lin Feng: Validation, Methodology, Investigation. Shiwei Gao: Writing – review & editing, Resources. Jinjin Xue: Writing – review & editing, Validation. Shengpeng Wang: Writing – review & editing, Validation. Xueping Wang: Writing – original draft. Fei Ye: Writing – review & editing, Validation. Anhui Gui: Writing – review & editing, Validation. Jing Teng: Writing – review & editing. Rui Luo: Writing – review & editing. Jia Chen: Writing – original draft. Zhonghua Liu: Writing – review & editing, Resources. Panpan Liu: Writing – review & editing, Writing – original draft, Resources, Funding acquisition, Conceptualization.

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.

Appendix A. Supplementary data

Supplementary material

Data availability

Data will be made available on request.