Using comprehensive two-dimensional liquid chromatography to unveil and mitigate the bidirectional quantification biases of non-volatile compounds in Chinese black tea processing Authors: Shanbo Zhang, Kim-Huey Ee, Yunle Huang, Rui Min Vivian Goh, Lingyi Li, Lionel Jublot, Chee Sian Gan, Shao Quan Liu, Bin Yu Published: 2026 ABSTRACT AbstractTo better understand Chinese black tea processing, the quantification of non-volatiles is essential, yet conventional one-dimensional liquid chromatography (1D-LC) struggles with increasing matrix complexity as fresh tea leaves undergo enzymatic oxidation and polymerization. This study evaluated comprehensive two-dimensional liquid chromatography (LC × LC) against 1D-LC for quantifying catechins and theaflavins across fresh tea leaves and four processing stages (withering, rolling, oxidation, and drying). Non-volatiles profiling with 1D-LC, revealed extensive co-elution across the chromatographic space, with unidentified features accumulating particularly in late-retention regions. The primary quantitative finding was that 1D-LC exhibited a severe bidirectional bias due to co-elution: catechins were overestimated (up to 4-fold), while theaflavins were underestimated (64%). Crucially, LC × LC mitigated these quantification errors by resolving theaflavins (TF3G and TF3’G) and minimizing matrix effects (ME), narrowing the ME range from 0.60 to 2.20 (1D-LC) to 0.75–1.25 (LC × LC). These results demonstrate LC × LC as a robust method for better quantification in complex food matrices.Keywords: Two-dimensional liquid chromatography, Quantification bias, Chinese black tea processing, Catechins, Matrix effect Introduction Black tea is particularly renowned for its rich flavor and diverse bioactive compounds (Chen et al., 2025). The processing of black tea involves a series of sequential processing steps—withering, rolling, oxidation (commonly referred to as fermentation), and drying—that modulate the chemical composition of tea leaves and ultimately determine product quality (Yang et al., 2024). In well-known Chinese black tea production, these processing techniques follow established steps and give rise to a wide range of high-quality products (Chen et al., 2021; Liu et al., 2023). Understanding these processing-induced chemical changes is essential for optimizing processing conditions, ensuring batch consistency, and elucidating the biochemical basis underlying black tea quality (Yu et al., 2020; Zhang et al., 2020). Therefore, state-of-the-art analytical methodologies have become indispensable for characterizing these chemicals, although significant challenges remain due to the continuously evolving and highly intricate nature of the tea matrix. To analyze these chemicals, liquid chromatography coupled with mass spectrometry (LC–MS)–based analytical approaches are commonly used (Cajka, 2024). For instance, LC–MS has been employed to monitor the transformation of polyphenols during Japanese black tea processing, revealing stage-specific changes in catechin dimers and flavonol glycosides (Ito & Yanase, 2022). Widely targeted LC–MS metabolomics has also been used to characterize dynamic variations in hundreds of flavonoids, phenolic acids, and amino-acid derivatives across the withering–rolling–fermentation–drying processing in the cultivar Camellia sinensis cv. Huangjinya (Zhou et al., 2022). These studies collectively demonstrate the utility of LC–MS for elucidating processing-related biochemical changes. Furthermore, LC-MS provides sufficient resolution and quantification ability for high-abundance components such as catechin monomers, particularly when optimized conditions are applied. However, as processing progresses, the tea matrix becomes increasingly complex due to enzymatic oxidation and polymerization reactions (Nasiri et al., 2021). In conventional one-dimensional LC-MS, the vast number of newly formed molecules often exceeds the peak capacity of the column, resulting in the co-elution of target analytes with matrix components (Plumb et al., 2023; Prabhu et al., 2023). This co-elution frequently leads to matrix effects—manifesting as signal suppression or enhancement in the MS source—which can cause biased abundance measurements (Hu et al., 2019; Qin et al., 2024). While numerous studies have quantified the dynamic changes of metabolites during tea processing (Liu et al., 2022; Peng et al., 2022), most relied on external solvent calibration, often neglecting the potential impact of matrix effects. Therefore, researchers are increasingly seeking new analytical strategies capable of handling the diversifying composition of tea during processing and mitigating the matrix-induced interferences that compromise quantitative accuracy. Two-dimensional liquid chromatography (2D-LC) has been increasingly applied to mitigate matrix-related interferences in complex food systems, offering improved separation selectivity compared with 1D-LC (Ning et al., 2025). Different operational modes exist within 2D-LC, each with distinct capabilities. Heart-cutting 2D-LC (LC–LC), for instance, enables higher selectivity but is inherently restricted to a limited number of preselected analytes because only discrete fractions are transferred to the second dimension (2D) (Steiner et al., 2020; Stoll & Carr, 2017). Selective comprehensive LC × LC (sLC × LC) extends better quantification to slightly broader analyte groups by employing intermittent modulation; however, its applicability remains constrained when the analytes span multiple chemical classes or exhibit widely varying abundances (Pirok et al., 2018; Zhang, Ee, et al., 2025). In contrast, fully comprehensive LC × LC continuously transfers the entire first-dimension effluent into the second dimension, generating a complete two-dimensional separation space with maximized chromatographic coverage (Pursch et al., 2017; Stoll & Carr, 2017). This configuration yields a broader and more dispersed distribution of compounds across the two-dimensional separation space, reducing co-elution between target compounds and interfering matrix constituents that typically drive ionization suppression or enhancement in mass spectrometry (Muehlwald et al., 2020). Moreover, the LC × LC separation provides a more complete visualization of the evolving chemical composition during tea processing, revealing minor oxidation products and isomeric species that remain unresolved in conventional one-dimensional chromatograms (Huang et al., 2022). LC × LC has also been widely validated as a quantitative analytical workflow for natural products. Multiple reviews and application studies have documented its successful use for the quantification of polyphenols, flavonoids, and other bioactive compounds in complex food and plant matrices (Lazzari et al., 2019; Lu et al., 2025). Although LC × LC has demonstrated clear advantages in resolving co-eluting compounds in complex food matrices, its application to tea processing studies remains scarce. This is partly because most previous investigations of tea chemistry have relied on one-dimensional LC–MS approaches aimed at qualitative profiling or relative comparison across processing stages, rather than rigorous absolute quantification (Shen et al., 2026). Under such study designs, chromatographic co-elution was often implicitly tolerated as long as global trends could be observed. To date, no study has systematically evaluated how matrix evolution during black tea processing compromises quantitative accuracy in 1D-LC, nor assessed whether comprehensive LC × LC can effectively resolve isomeric theaflavins and mitigate bidirectional quantification bias across processing stages (Caño-Carrillo et al., 2024). In this study, a comprehensive two-dimensional liquid chromatography coupled with quadrupole-time-of-flight mass spectrometry (LC × LC–QTOF) approach was applied to characterize the chemical transformations occurring throughout Chinese black tea processing, with particular focus on major compounds such as catechins and theaflavins. The objective of this study was to systematically evaluate the impact of matrix evolution during Chinese black tea processing on the quantitative reliability of conventional one-dimensional LC–MS, and to assess whether comprehensive two-dimensional liquid chromatography (LC × LC–QTOF) can effectively resolve co-eluting compounds, mitigate matrix-induced ionization bias, and enable reliable quantification of catechins and theaflavin isomers across different processing stages. Materials and methods Chemicals and reagents Catechins and theaflavins were selected because they represent some of the most abundant bioactive compounds in black tea (Aydemir et al., 2024). To analyze these compounds, analytical standards (>98% purity) including catechin (C), epicatechin (EC), gallocatechin (GC), epigallocatechin (EGC), epicatechin-3-gallate (ECG), epigallocatechin-3-gallate (EGCG), theaflavin (TF), theaflavin-3-gallate (TF3G), theaflavin-3′-gallate (TF3’G), theaflavin-3,3′-digallate (TFDG) were purchased from FineTech Industry Limited (Wuhan, Hubei, China). The stock solution of each standard was dissolved in pure water and stored at 4 °C. LC-MS grade water, methanol (MeOH), and acetonitrile (ACN) were purchased from Thermo Fisher Scientific (Waltham, Massachusetts, USA), while LC-MS grade formic acid (FA) was obtained from Merck (Darmstadt, Germany). Sample collection and sample preparation Sample collection Fresh tea leaves (Camellia sinensis var. sinensis), each with one bud and two leaves, were harvested from Wuyishan County, Nanping City, Fujian Province in April 2025. The harvested fresh tea leaves (FL) were collected and subjected to traditional black tea (BT) processing, which included four major processing steps: withering (WT), rolling (RL), fermentation/oxidation (OX), and drying (DR). The details of the processing conditions are provided in Table S1. At the end of each processing step, three biological replicates (∼200 g each) were collected. The samples were steamed and then dried at 60 ± 2 °C for 1 h, subsequently sealed in airtight polyethylene bags, and snap-frozen on dry ice. All samples were transported to the laboratory on dry ice and subsequently stored at −80 °C until further analysis. Sample preparation Prior to extraction, the precise water content of the processed tea samples was determined to ensure accurate quantification. The moisture content was calculated by comparing the weight of the samples before and after the drying process until a constant dry weight was achieved. All analytical data and compound concentrations were subsequently normalized and expressed on a dry weight (DW) basis to eliminate potential bias caused by moisture variations across different processing stages. For extraction, lyophilized tea samples from each processing stage were ground into fine powder using a laboratory mill (30 s, twice) and passed through a 60-mesh sieve (0.25 mm) to ensure particle size uniformity. Subsequently, 1.00 g ± 0.05 g (DW) of the powdered sample was used for extraction. Extraction method was based on previous described methods (Zhang, Long, et al., 2025). Each sample was extracted with 20.0 mL of acetone/water (50:50, v/v) at 50 °C by sonication in a water bath for 123 min. This specific solvent-to-sample ratio and extraction duration were adopted based on a prior optimization study to maximize the yield of catechin and catechin derivatives. After extraction, the extracts were centrifuged at 12,000 ×g at 4 °C for 10 min. Then the supernatant was collected and filtered through a 0.22-μm PTFE membrane (Ctech Global Pte Ltd., Singapore) prior to liquid chromatography-quadrupole-time-of-flight mass spectrometer (LC–QTOF) analysis. To avoid the solvent effect of acetone and ensure the signal intensity remained within the linear range of the QTOF detector, the supernatant was diluted 10–1000 times with LC mobile phase. 1D-LC analysis An Agilent 1290 Infinity II ultra-high performance liquid chromatography (UHPLC) system coupled with a 6550 iFunnel QTOF was used (Agilent Technologies, Santa Clara, California, USA) for analysis. For 1D-LC separation, a Poroshell HPH-C18 column (2.1 × 150 mm, 2.7 μm; Agilent Technologies, CA, USA) was used. The elution program was adapted from Huang et al. (2022) with modifications to improve resolution on the HPH-C18 column and to generate appropriate fraction volumes for subsequent LC × LC analysis. The mobile phases consisted of (A) water containing 0.1% formic acid and (B) MeOH. Separation was initiated at 5% solvent B and held for 3 min, followed by a continuous multi-step linear gradient in which solvent B was increased to 16% over 12 min, to 18.5% over the subsequent 5 min, to 21% over the next 3 min, to 23.5% over 5 min, and then to 50% over 10 min. The organic composition was subsequently increased to 95% B within 5 min and maintained for an additional 5 min to wash the column, after which the system was returned to the initial conditions for re-equilibration. The flow rate was set at 0.12 mL/min, with the column temperature maintained at 40 °C and an injection volume of 1 μL. LC × LC analysis For the LC × LC separation, the first dimension employed the same HPH-C18 column as that used in 1D-LC, while the 2D utilized a Poroshell Bonus-RP column (3.0 × 50 mm, 1.9 μm; Agilent Technologies, CA, USA). The modulation was performed using a dual-loop active modulation interface, in which two 60 μL sampling loops operated in an alternating manner. While one loop collected the first-dimension effluent, the other was simultaneously injected into the second dimension, thereby enabling continuous and comprehensive modulation. A modulation period of 0.25 min was applied to LC × LC analysis. The 2D separation was operated using a series of short, fast gradients that were adjusted as a function of the first-dimension (1D) elution time to maximize orthogonality. During the initial 0–5 min of the 1D separation, the 2D gradient was initiated at 28% solvent B and rapidly increased to 40% B within 0.2 min. As the 1D gradient progressed, the initial organic composition of the 2D gradient was increased to 30% B between 5 and 20 min and ramped to 50% B within 0.2 min. During the later elution window of the 1D (20–40 min), the 2D gradient started at 40% B and was increased to 60% B within 0.2 min. After 40 min of the 1D separation, the 2D gradient conditions were kept constant for the remainder of the analysis. The mobile phase flow rate was set to 2.18 mL/min, and the column temperature was maintained at 40 °C. Detection A QTOF detector was used for both 1D-LC and LC × LC analyses. The QTOF was operated under negative electrospray ionization (ESI) mode. Source conditions were as follows: gas temperature, 200 °C; drying gas flow, 16 L/min; nebulizer pressure, 35 psi; sheath gas temperature, 350 °C; sheath gas flow, 11 L/min; capillary voltage, 2500 V; and nozzle voltage, 0 V. Mass spectra were acquired over an m/z range of 50–1500 at a rate of 8 spectra/s. Mass accuracy was maintained using a reference solution according to manufacturer's guidelines, and signal consistency was monitored with pooled quality control samples. All analyses were conducted in triplicate to ensure reproducibility. Characteristic [M–H]− adducts of the major catechins (e.g. EC/C, m/z 289.0718; EGC/GC, m/z 305.0667; ECG/CG, m/z 441.0827; EGCG/GCG, m/z 457.0776) and theaflavins (e.g. TF, 563.1195, m/z; TF3G, 715.1305, m/z; TF3’G, 715.1305, m/z; TFDG, 867.1414, m/z) were used for identification. Method validation and sample analysis Linearity and sensitivity The linearity of the analytical method was evaluated by analyzing pure standard solutions at eight different concentration levels ranging from 0.005 to 10 μg/mL. Stock solutions for each standard were prepared in MeOH at a concentration of 1000 μg/mL. All samples were appropriately diluted prior to analysis, to suit the calibration ranges. Calibration curves were constructed by plotting the peak area (for 1D-LC) or 2D blob area (for LC × LC) against the corresponding analyte concentration. In LC × LC analysis, a “2D blob” was defined as the contiguous signal region corresponding to a given analyte in the two-dimensional separation space, as visualized in the LC × LC contour plot. Blob boundaries were determined based on consistent first- and second-dimension retention time windows established from repeated injections of pure standards. Integration of the 2D blob area was performed using software-assisted two-dimensional integration, in which the total signal intensity within the defined retention time region was summed. The same integration criteria were applied consistently to both calibration standards and sample analyses (Pérez-Cova et al., 2022). The calibration range was adjusted specifically for each analyte based on its limit of quantification (LOQ). The limits of detection (LOD) and LOQ were determined to assess the sensitivity of the method. LOD and LOQ were determined as the concentrations of analytes yielding signal-to-noise ratios of 3-to-1 and 10-to-1, respectively (Cantwell, 2025). Assessment of matrix effects To evaluate potential matrix effects (MEs), two different calibration curves were constructed for each analyte under the applied analytical conditions. One curve was generated using pure standards dissolved in solvent (unspiked), while the other was obtained by spiking the sample extracts with the same standards at equivalent concentration levels (spiked). For matrix effect evaluation, A was defined as the slope of the calibration curve obtained from standards prepared in pure solvent (unspiked), while B corresponded to the slope of the calibration curve obtained from standards spiked into black tea extracts after extraction. The matrix effect was then quantified based on the ratio of these two slopes according to the method described by Gosetti et al. (2010), as expressed by the following equation: ME (%) = BA×100. Precision and accuracy The precision of the method was evaluated by determining the relative standard deviation (%RSD) of peak areas at an intermediate concentration level (0.25 μg/mL). Intra-day precision was assessed through triplicate injections performed within a single day, while inter-day precision was determined by analyzing the samples across three consecutive days (n = 9). Accuracy was also evaluated alongside precision to ensure the reliability of the analytical method. Data processing and statistical analysis MassHunter (Version 12.0, Agilent, CA, USA) was used to perform LC–QTOF data acquisition. MassHunter Qualitative Analysis software (Version 12.0, Agilent, CA, USA) was used to perform raw data processing. Multidimensional chromatographic data were processed using LC Image™ software (GC Image, LCC and the University of Nebraska; version 7.2e2.2, LC × LC-HRMS, 2001–2017). LC × LC data consist of a series of modulated 2D chromatograms generated at defined fractions of the 1D. Within LC Image™, LC × LC peak detection was conducted by first applying automated peak-picking to each 2D slice to identify retention time features in the 2D, followed by clustering of these features across consecutive slices to reconstruct distinct LC × LC peaks spanning both first- and second-dimension retention spaces. This step includes baseline correction, noise filtering, and threshold-based peak recognition to discriminate true chromatographic signals from background. Next, detected LC × LC peaks were aligned across all samples by matching retention times in both dimensions (using retention time tolerances and pattern similarity across runs), allowing construction of a consistent feature matrix for comparative analysis. Peak integration was then performed by summing signal intensities over the defined LC × LC peak regions (accounting for both dimensions) to generate quantitative peak volumes for each feature across samples. Data filtering and quality improvement were carried out using Agilent MassHunter Explorer (version 1.0). Statistical analyses were performed with OriginPro (version 2022; OriginLab Corporation, Northampton, MA, USA) at a 95% confidence level. All results are reported as mean ± standard deviation based on triplicate analyses. Results and discussion Co-eluting challenges across different black tea processing stages Black tea processing involves a complex series of biochemical and chemical transformations that fundamentally alter the chemical composition of fresh tea leaves. To comprehensively characterize these changes, a conventional LC-QTOF profiling across four sequential processing stages together with fresh tea leaves (FL): withering (WT), rolling (RL), oxidation (OX), and drying (DR) was carried out. The raw data were pre-processed, yielding a dataset of 3617 ion features. Each ion feature represents a unique combination of accurate mass to charge ratio (m/z) and retention time, corresponding to a distinct chemical entity in the tea matrix. Among these detected features, 74 compounds were putatively identified through accurate mass matching (<10 ppm error) and MS/MS fragmentation pattern comparison against reference databases and authentic standards (Table S2). The identified compounds spanned multiple chemical classes: 12 amino acids (theanine, arginine, lysine, histidine, phenylalanine, and tryptophan, etc.), 3 alkaloids (caffeine, theobromine, and theophylline), 18 flavan-3-ols and their oxidation products (catechin monomers, procyanidins, theaflavins, and theasinensins), 25 flavonol glycosides (quercetin, kaempferol, and myricetin derivatives), 12 phenolic acids (caffeoylquinic acids, feruloylquinic acids, and gallic acid), and 3 sugars (D-Mannose, D-Sucrose, and D-Maltose). These compounds serve as anchor points within the highly complex chromatograms. Although 74 compounds were putatively identified (≈2% of 3617 features), they provide a useful framework to illustrate how targeted analytes are affected by co-elution within the tea matrix. To exemplify this, extracted ion chromatograms (EICs) of two representative compounds, EGCG and TFDG, were examined across the processing stages (Fig. S1). These EICs show changes in abundance during processing; however, each apparent single EIC peak commonly encompasses numerous co-eluting ion features that are invisible in a single-target trace. For example, the peak region of EGCG contains relatively modest changes in the number and intensity of background ion features, whereas the retention window of TFDG exhibits a progressive increase in co-eluting ion features as tea processing advances. These unidentified features likely include catechin oxidation intermediates (such as quinones and semiquinones), oligomeric procyanidins of varying chain lengths, heterogeneous thearubigin polymers, Maillard reaction products (such as melanoidins, furan derivatives and pyrrole compounds), and countless other transformation products generated during enzymatic oxidation and thermal processing (Nasiri et al., 2021; Utpott et al., 2022). As revealed above, a prominent analytical challenge became immediately apparent as numerous compounds exhibited overlapping retention times in 1D-LC, creating severe chromatographic congestions. To systematically illustrate the scope of unidentified features co-eluting with target analytes, six representative retention time regions (Regions 1–6 (R1-R6)) were selected based on the elution windows of major catechins and theaflavins. Among the identified compounds (Table 1), catechin (C) and gallocatechin (GC) both eluted at RT ∼18.900 min as depicted in R2, while all four theaflavins clustered within a narrow 0.200-min window at RT 42.900–43.100 min as depicted in R6. This co-elution extended beyond structurally similar isomers; compounds from entirely different chemical classes also converged at similar retention times. For example, in the retention window surrounding theaflavins (R6, RT 42.500–43.500 min), Table 1 reveals co-elution with eight structurally unrelated compounds including flavonol glycosides (rutin, naringin), flavanones (naringenin), flavone derivatives (luteolin-7-O-rhamnoside) and three unidentified features (m/z 1033.2889, 871.2361, 1175.5562). After LC × LC separation, the second dimension (RT₂) values in Table 1 demonstrate that LC × LC expanded the effective chromatographic space by introducing a different selectivity dimension. Some co-eluting clusters that were inseparable in 1D-LC were spread out across the 2D. Table 1Putative identified compounds and unidentified ion feature co-eluting in six retention time regions.Co-elution region (1D)Compound/featureRT1a(min)RT2b(s)AdductMeasuered m/zIdentificationcRegion 1 (R1)Unknown feature 0111.98614.2[M-H]-239.1061MSEpigallocatechin (EGC)12.0032.4[M-H]-305.0666StdRhoifolin12.0561.9[M-H]-577.1561MS/MSEpiafzelechin 3-O-gallate12.0692.5[M-H]-425.0878MS/MSUnknown feature 0212.0773.3[M-H]-411.0274MSUnknown feature 0312.0773.3[M-H]-315.0745MSRegion 2 (R2)Hyperoside18.9216.3[M-H]-463.0882MS/MSCatechin (C)18.9334.7[M-H]-289.0725StdGallocatechin (GC)18.9332.9[M-H]-305.0671StdUnknown feature 0418.9445.7[M-H]-697.1803MSUnknown feature 0518.9446.3[M-H]-405.0838MS3,4-di-O-caffeoylquinic acid18.9497.5[M-H]-529.1351MS/MS4-O-Caffeoyl shikimic acid18.9564.5[M-H]-335.0772MS/MSRegion 3 (R3)Epigallocatechin gallate (EGCG)25.2414.5[M-H]-457.0778Std4-p-Coumaroylquinic acid25.2417.5[M-H]-337.0929MS/MSEriocitrin25.2416.3[M-H]-595.1668MS/MSUnknown feature 0625.25510.1[M-H]-525.0691MSUnknown feature 0725.26112.1[M-H]-555.0990MSMyricetin 3-O-glucoside25.2764.9[M-H]-479.0836MS/MSRegion 4 (R4)Unknown feature 0827.3851.8[M-H]-482.0839MSUnknown feature 0927.3882.1[M-H]-581.1557MSEpicatechin (EC)27.4082.4[M-H]-289.0724StdEriodictyol-7-glucoside27.4561.9[M-H]-449.1089MS/MSOolongtheanine27.4983.4[M-H]-731.1254MS/MSRegion 5 (R5)Unknown feature 1034.2664.1[M-H]-540.0045MSEpicatechingallate (ECG)34.2764.9[M-H]-441.0827StdOolongtheanin 3′-gallate34.2904.3[M-H]-883.1363MS/MS3,4-di-O-caffeoylquinic acid34.3064.4[M-H]-515.1195MS/MSVicenin-234.3394.5[M-H]-593.1515MS/MSUnknown feature 1134.3455.1[M-H]-883.1792MSRegion 6 (R6)Theaflavin (TF)42.89311.7[M-H]-563.1199StdTheaflavin-3,3′-digallate (TFDG)42.89312.3[M-H]-867.1430StdRutin42.9885.3[M-H]-609.1461MS/MSNaringenin43.0116.2[M-H]-271.0612MS/MSTheaflavin-3-gallate (TF3G)43.0237.8[M-H]-715.1311StdTheaflavin-3′-gallate (TF3’G)43.0236.9[M-H]-715.1314StdLuteolin-7-O-rhamnoside43.0446.8[M-H]-593.1512MS/MSNaringin43.0658.3[M-H]-579.1719MS/MSUnknown feature 1243.1127.4[M-H]-1033.2889MSUnknown feature 1343.1267.9[M-H]-871.2361MSUnknown feature 1443.1338.2[M-H]-1175.5562MSaRT1 = 1D retention time.bRT2 = 2D retention time.cIdentification methods: “Std”, confirmed using authentic reference standards; “MS”, identified based on accurate mass; “MS/MS”, identified by matching mass spectra fragmentation patterns with databases. Putative identified compounds and unidentified ion feature co-eluting in six retention time regions. RT1 = 1D retention time. RT2 = 2D retention time. Identification methods: “Std”, confirmed using authentic reference standards; “MS”, identified based on accurate mass; “MS/MS”, identified by matching mass spectra fragmentation patterns with databases. These observations underscore the intrinsic limitations of 1D-LC for highly matrix-evolving systems like black tea and encourage the application of LC × LC to decouple co-eluting species and mitigate matrix-induced quantification biases. Improved chromatographic resolution by LC × LC To address the severe co-elution challenges and dynamic matrix complexity, a LC × LC strategy employing different selectivity columns was implemented. The LC × LC system coupled two chromatographic dimensions with complementary selectivity: the first dimension (1D) utilized a reversed-phase C18 column providing hydrophobicity-based separation identical to conventional 1D-LC, while the 2D employed a Bonus-RP stationary phase that would introduce additional selectivity based on π-π interactions, shape recognition, and steric factors. Fig. 1 presents a comparison of 1D-LC and LC × LC separation performance across all four processing stages and fresh tea leaves (FL, WT, RL, OX, DR). In the 1D-LC total ion chromatograms (TIC), the fresh leaves exhibited a relatively simple chromatographic profile with discrete peaks. However, by the OX and DR stages, severe chromatographic congestions and elevated baseline emerged, particularly in the late-elution region (RT > 35.000 min) where theaflavins eluted. The improvement was apparent from the LC × LC contour plot: peaks that severely overlapped in 1D-LC were now dispersed across two-dimensional space, with each analyte occupying a unique (RT1, RT2) position. The improvement was particularly striking in later black tea processing stages (OX and DR), where 1D-LC showed a maximum congestion but LC × LC maintained excellent resolution with discrete, well-separated spots. Comparison of chromatographic separation between 1D-LC and LC × LC for selected catechins and theaflavins in black tea. Total ion chromatograms from 1D-LC (left) and contour plots from LC × LC (right) are shown for fresh tea leaves (FL) and four black tea processing stages (Withering, WT; Rolling, RL; Fermentation, OX; Drying, DR). The light blue boxes labeled R1–R6 indicate six representative retention time regions selected to illustrate severe co-elution challenges; the detailed list of identified analytes and co-eluting interferences within each region is provided in Table 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) The theaflavin retention region (R6) exemplified this improvement. In 1D-LC, all four theaflavins (TF, TFDG, TF3G, TF3’G) clustered within a narrow 0.200-min window at RT 42.900–43.100 min. In LC × LC, these four compounds were fully separated across the 2D, each appearing as a distinct spot. Critically, TF3G and TF3’G—positional isomers completely indistinguishable in 1D-LC—were successfully resolved in the 2D, enabling their better quantification. Some co-eluting compounds: (1) In R2, C and GC, which co-eluted at RT1 = 18.933 min in 1D-LC, were differentiated in 2D with RT2 = 4.7 s and 2.9 s, respectively. (2) In R6, four theaflavins, clustered within RT1 42.893–43.023 min (0.130-min window) in the first dimension, spanned RT2 6.9–12.3 s (5.4-s range) in the 2D, achieving better separation and (3) TF3G (RT2 = 7.8 s) and TF3’G (RT2 = 6.9 s), which shared identical RT1 = 43.023 min and were unquantifiable in 1D-LC, were now successfully separated, enabling calculation of the monogallate-to-digallate theaflavin ratio—an important quality marker correlating with tea brightness and astringency (Leung et al., 2001). Furthermore, interferences identified in Table 1, such as Rutin (RT2 = 5.3 s) and Naringin (RT2 = 8.3 s), are shifted away from the theaflavin quantitation spots, ensuring spectral purity. Evaluation of matrix effects between LC × LC and 1D-LC While the findings presented in Section 3.2 demonstrated that LC × LC markedly improved chromatographic resolution—particularly in later black tea processing stages where severe co-elution occurred—the analytical implications of this improvement extend beyond peak dispersion alone. Black tea processing fundamentally reshapes the chemical matrix: catechins diminish, theaflavins and thearubigins accumulate, and polar amino acids, sugars, and semi-polar non-volatile compounds fluctuate across stages. The representative 3D LC × LC contour plot further illustrates the intrinsic complexity of the black tea matrix, showing dense background features and multiple co-eluting regions that would be indistinguishable under 1D-LC conditions (Fig. S2). In a 1D-LC system, these dynamic matrix constituents frequently co-elute with target analytes, causing insufficient separation that translates directly into matrix effect enhancement or suppression in ESI-MS and ultimately leads to biased quantitative trends. Therefore, this Section evaluates matrix effects (MEs) across processing stages and between methods to determine whether LC × LC reduces such interferences. Fig. 2 illustrates the MEs observed for selected analytes—specifically, the EGCG, GCG, EC, ECG, and the TF and TFDG—across sequential stages of black tea processing: FL, WT, RL, OX, and DR. The X-axis represents these processing stages, while the Y-axis quantifies the ME as a dimensionless ratio, calculated according to the formula ME = (peak area of analyte in spiked sample matrix) / (peak area of analyte in neat solvent standard), where a value of 1 signifies negligible matrix-induced interference, values <1 denote ion suppression (leading to analyte signal underestimation and potential quantification bias), and values >1 indicate ion enhancement (resulting in signal overestimation). Determinations were performed using either 1D-LC (black bars) or LC × LC (orange bars), with error bars depicting the standard deviation from triplicate analyses (n = 3). Statistical comparisons between methods at each stage are marked with asterisks above the bars (* p < 0.05, ** p < 0.01, *** p < 0.001; ns = not significant, via unpaired two-tailed Student's t-test). Comparison of matrix effects between 1D-LC and LC × LC for selected catechins and theaflavins across fresh tea leaves and four black tea processing stages (FL, WT, RL, OX, DR). Data represent mean ± SD (n = 3). * p < 0.05, ** p < 0.01, *** p < 0.001, and ns = not significant. Black tea processing involves progressive chemical and physical transformations of tea leaves. Catechins (flavan-3-ols) undergo enzymatic oxidation by polyphenol oxidase (PPO) to form theaflavins (benzotropolone structures) and, ultimately, thearubigins (high-molecular-weight polymers) (Chen et al., 2012). These transformations increase matrix complexity by generating diverse phenolic oxidation products. Additionally, tea leaves inherently contain polar and semi-polar compounds that can co-elute or compete for ionization in LC–MS, further exacerbating MEs (Ding et al., 2002). From Fig. 2, it can be observed that the limited separation power in 1D-LC, often resulted in co-elution of these interferents with target analytes, resulting in ME ranges of 0.6–2.2 across compounds and processing stages. In contrast, LC × LC employed improved separation powers, which separate analytes from matrix components and confines MEs to a tighter 0.80–1.25 range. Across all stages, LC × LC consistently outperformed 1D-LC (p < 0.01 in >80% of cases), facilitating precise tracking of compound conversions and their bioavailability-relevant concentrations for health-relevant applications, such as antioxidant capacity assessments. Fresh leaves (FL): As unprocessed Camellia sinensis tea leaves, the matrix is biochemically simplest, dominated by native catechins (up to 20–30% dry weight) with minimal oxidation products or degradants (Qiu et al., 2024). Consequently, MEs for all analytes hover near unity in both methods (e.g., ME (EC) in 1D-LC = 0.98, ME (EC) in LC × LC = 1.02; ns), reflecting low interferent load and baseline compatibility for LC-MS. Minor deviations, if any (e.g., slight enhancement for TF precursors, though negligible here), arose from endogenous chlorophyll or flavonoids, but LC × LC offered no added benefit at this nascent stage. Withered leaves (WT): Withering involves controlled dehydration (37.679% moisture loss after 16 h withering), which concentrates the leaf's soluble constituents while initiating subtle enzymatic transformation. This process elevates the relative abundance of polar matrix components, such as free amino acids (e.g., L-theanine, comprising up to 1–2% dry weight) and simple carbohydrates, without substantial oxidation yet (Qiao et al., 2023). In electrospray ionization (ESI) mass spectrometry—commonly operated in negative mode for catechins—these concentrated polar neutrals can promote ion enhancement by facilitating more efficient charged droplet formation and desolvation in the ESI plume (Gosetti et al., 2010; Liigand et al., 2021). This results in 1D-LC ME values subtly exceeding 1 for many catechins (e.g., ME (GC) = 2.0, p < 0.01 vs. FL), reflecting signal overestimation due to favorable matrix-analyte interactions. In contrast, LC × LC mitigated such enhancement through improved separation, diluting co-eluting enhancers and yielding MEs closer to 1, thus preventing over-quantification in early dynamic profiling. This stage-specific improvement underscores the need for dimensionality in separation to normalize subtle matrix modulations arising from pre-oxidative concentration effects. Rolled leaves (RL): Rolling applies mechanical shear to withered leaves, fracturing vacuolar membranes and promoting the efflux of polyphenol oxidase (PPO) enzymes alongside catechin substrates into the extracellular space, thereby catalyzing localized oxidative coupling and the de novo synthesis of theaflavin precursors, thereby increasing matrix complexity (Chen et al., 2023). At this stage, 1D-LC ME values often deviate significantly (e.g., GC and EGC showing ion enhancement), likely due to co-elution of early oxidation products and other co-existing non-volatile compounds that would amplify the signal. Conversely, theaflavins, which emerge later at RT1 = 42.000–43.000 min (e.g., ME (TF) = 0.82, ME (TFDG) = 0.79, * vs. WT), were now detectable as oxidation markers. These compounds experienced ion suppression due to competitive adduction by co-migrating hydrophobic interferents, such as catechin dimers or trimers, which would sequester ESI charges and reduce signal intensity by 18–21%, potentially leading to an underestimation of catechin-to-theaflavin conversion efficiencies. LC × LC efficiently separated these interfering species, restoring ME values closer to 1, demonstrating its advantage under conditions of enzyme-induced matrix complexity. Oxidized/Fermented leaves (OX): During fermentation, PPO-driven catechin oxidation generates the highest theaflavin levels alongside abundant thearubigin polymers (>20%), which accumulated in the late-elution portion of the chromatogram (RT1 > 35.000 min) and increased matrix complexity (Aloo et al., 2024). In 1D-LC, residual catechin monomers, which elute earlier (RT1 = 12.000–34.000 min; e.g., ME (EGCG) = 0.92, ns vs. LC × LC), exhibited comparable MEs (∼0.90–1.10) across both 1D-LC and LC × LC. This is because their higher polarity minimizes co-elution with late-eluting polymeric suppressors, while many small, ion-enhancing species such as sugars, organic acids, and other non-volatile compounds—abundant during the RL—have been consumed or transformed during fermentation, mitigating the previously observed ion enhancement. In contrast, theaflavins (RT1 = 42.000–43.000 min; e.g., ME (TF) = 0.74, ***) suffered substantial ion suppression (25–30% signal loss) from charge-competitive humic and quinone adducts. LC × LC effectively resolved these late-eluting suppressors via enhanced separation, restoring theaflavin MEs (ME (TF) = 1.22, ***). These results highlight the stage- and elution-time-specific selectivity of LC × LC, enabling accurate quantification of both early catechins and late-forming theaflavins, and revealing >80% conversion efficiencies relevant to bioactivity optimization. Dried leaves (DR): In the final drying stage (DR), high-temperature treatment halts enzymatic oxidation and stabilizes polyphenolic products (Topal & Şahi˙n, 2025). In 1D-LC, MEs persisted for late-eluting digallate theaflavins (ME (TFDG) = 0.75, *) due to hydrophobic co-retention and ion suppression by residual degradation products. By the DR, early-eluting catechins no longer exhibited ion enhancement, but instead showed mild ion suppression (e.g., ME (EC) = 0.95–0.98). This shift reflects the depletion of small, polar compounds that previously contributed to signal enhancement during rolling and fermentation. Concurrently, residual polyphenolic polymers and thermal degradation products introduced low-level competitive suppression in the ESI source. The combined effect of reduced enhancing species and persistent suppressors resulted in a slight net decrease in ionization efficiency for early-eluting catechins. LC × LC, however, effectively separated these interfering species, normalizing ME (TFDG) to 0.96 and maintaining ME values between 0.75 and 1.1 for all monitored compounds. In summary, matrix complexity progressively increases throughout black tea processing, leading to variable ME patterns across compounds in 1D-LC, often showing initial enhancement followed by suppression. In contrast, LC × LC maintains ME values within a narrower range (0.75–1.25), substantially mitigating interferences. This finding directly demonstrates that 1D-LC is quantitatively insufficient for this application, and that LC × LC provides the necessary resolving power to obtain accurate, matrix-corrected concentrations. These results not only validate the anti-interference capability of LC × LC but also provide reliable data for tracking the loss and transformation of health-relevant polyphenols during black tea processing steps. Quantification accuracy and implications for Chinese black tea processing Before conducting quantitative analysis of tea samples, the analytical performance of both LC-QTOF and LC × LC-QTOF methods was rigorously validated using pure standards of eight catechin monomers and four theaflavins prepared in blank matrix (Table 2). Validation parameters included linearity, limits of detection (LOD) and quantification (LOQ), and precision (intra- and inter-day repeatability). Table 2Analytical performance of catechins and theaflavins using 1D-LC and LC × LC.CompoundLinear range (μg/mL)R2LOD (μg/mL)LOQ (μg/mL)Intra-day RSD% (n = 3)Inter-day RSD% (n = 9)1D-LCLC × LC1D-LCLC × LC1D-LCLC × LC1D-LCLC × LC1D-LCLC × LC1D-LCLC × LCEpigallocatechin (EGC)0.0100–10.00000.0100–10.00000.99980.99990.00210.00210.00730.00790.510.474.683.91Catechin (C)0.0050–2.00000.0100–2.00000.99890.99970.00130.00230.00470.00870.970.564.171.02Gallocatechin (GC)0.0200–10.00000.0500–10.00000.99980.99980.00570.01350.01700.04300.990.844.112.66Epigallocatechin gallate (EGCG)0.0200–10.00000.0200–10.00000.99980.99950.00570.00570.01600.01601.390.950.251.48Epicatechin (EC)0.0050–5.00000.0100–5.00000.99970.99990.00110.00210.00380.00681.230.994.132.78Epicatechingallate (ECG)0.0050–10.00000.0100–10.00000.99870.99900.00140.00340.00460.00930.331.093.161.62Theaflavin (TF)0.0500–10.00000.1000–10.00000.99990.99890.01400.02810.04300.08902.332.101.653.41Theaflavin-3,3′-digallate (TFDG)0.0500–10.00000.1000–10.00000.99890.99890.01400.03030.00410.09731.090.564.165.71Theaflavin-3-gallate (TF3G)–0.1000–10.0000–0.9976–0.0322–0.0991–2.09–2.33Theaflavin-3′-gallate (TF3’G)–0.1000–10.0000–0.9995–0.0320–0.0990–0.55–2.05“-” indicates compounds that were not separated and therefore were not considered for quantification in 1D-LC. Analytical performance of catechins and theaflavins using 1D-LC and LC × LC. “-” indicates compounds that were not separated and therefore were not considered for quantification in 1D-LC. Both methods exhibited excellent linearity across the tested concentration ranges, with correlation coefficients (R2) ≥ 0.9976 for all analytes, indicating robust quantitative performance. The linear dynamic ranges spanned 2–3 orders of magnitude, covering expected physiological concentrations of polyphenols in tea extracts. For catechins, 1D-LC showed lower LODs (0.0011–0.0057 μg/mL) compared to LC × LC (0.0021–0.0135 μg/mL), likely reflecting the absence of sample transfer losses in single-dimension separation. For theaflavins, LC × LC achieved comparable sensitivity (LODs 0.0281–0.0322 μg/mL), with the critical advantage of enabling quantification of TF3G and TF3’G—compounds that were undetectable in 1D-LC due to severe co-elution with flavonol glycosides in R6 (Table 1, RT 42.900–43.100 min). Precision was evaluated through intra-day (n = 3) and inter-day (n = 9) repeatability, with relative standard deviations (RSD%) ≤ 2.33% (intra-day) and ≤ 5.71% (inter-day) for both methods, demonstrating high reproducibility. Importantly, LC × LC improved precision for the majority of compounds. For example, inter-day RSDs for C, EC, and GC decreased substantially in LC × LC relative to 1D-LC (C: 4.17% → 1.02%; EC: 4.13% → 2.78%; GC: 4.11% → 2.66%), reflecting reduced variability from co-eluting interferents. Intra-day precision also showed notable improvements for most analytes (e.g., C: 0.97% → 0.56%; EGCG: 1.39% → 0.95%). Slightly higher inter-day RSDs were observed for some late-eluting theaflavins (e.g., TFDG: 5.71% vs. 4.16% in 1D-LC), likely reflecting day-to-day fluctuations in modulation timing rather than limitations of the separation itself. Overall, Table 2 demonstrates that both methods meet rigorous analytical criteria; however, LC × LC provides consistently improved quantitative performance for most analytes especially the ability to quantify TF3G and TF3’G that were not previously available using 1D-LC approach. While superior resolution of isomeric theaflavins (TF3G and TF3’G) is a major advantage, the improved precision observed for catechins and several theaflavins cannot be explained by isomer separation alone. Instead, these improvements align with the reduced matrix effects quantified in Section 3.3, where LC × LC confined ME values to a narrower 0.80–1.25 range compared with the broader 0.60–2.20 range observed in 1D-LC. The mitigation of matrix-induced ion suppression and enhancement directly contributes to the more stable RSDs reported in Table 2. Together, the improved selectivity, reduced ME bias, and reliable precision establish LC × LC as a more robust quantitative platform for tracking metabolite transformations during black tea processing. Fig. 3 presents a comprehensive comparison of catechin and theaflavin quantification data across four black tea processing stages and fresh tea leaves (FL, WT, RL, OX, and DR) using 1D-LC (blue bars) and LC × LC (orange bars) methods. Statistical analysis (unpaired Student's t-test) revealed significant differences between the two methods for the majority of analytes across the tea processing stages. In general, the dominant esterified catechins (EGCG and ECG) and EC exhibited progressive decreases from fresh leaves to finished black tea, consistent with oxidative polymerization during processing. In contrast, non-esterified catechins (EGC, GC, and C) showed an initial accumulation during the withering and rolling stages before declining significantly during oxidation. For instance, EGCG, the most abundant catechin, decreased from approximately 26.0–28.0 mg/g in FL to ∼5.0 mg/g after drying. Conversely, theaflavins (TF and TFDG) accumulated significantly during OX and DR, with TFDG reaching maximum concentrations at the DR stage (1D-LC: 1.2 mg/g vs. LC × LC: 3.3 mg/g). Comparison of quantitative results obtained from 1D-LC and LC × LC for selected catechins and theaflavins across fresh tea leaves and four black tea processing stages (FL, WT, RL, OX, DR). Concentrations (mg/g DW) of selected catechin and theaflavin compounds measured by 1D-LC (blue) and LC × LC (orange). Data represent mean ± SD (n = 3). * p < 0.05, ** p < 0.01, *** p < 0.001, and ns = not significant. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) The 1D-LC and LC × LC methods exhibited contrasting quantification patterns between catechins and theaflavins throughout the black tea processing stages. As the matrix effect (ME) in LC × LC was close to 1 for all target compounds, the values obtained from LC × LC suggest more consistent measurements under complex matrix conditions. For the six catechin monomers, 1D-LC overestimated concentrations relative to LC × LC across most processing stages. In FL, EC showed relatively close agreement between the two methods (1D-LC: 24.7 mg/g; LC × LC: 25.1 mg/g; no significant difference), but this discrepancy became more pronounced during subsequent stages. At the WT stage, EC measurements diverged notably (1D-LC: 26.2 mg/g; LC × LC: 20.9 mg/g; significant difference), representing a 26% overestimation by 1D-LC. The most substantial overestimation occurred with GC at the WT stage, where 1D-LC measured 6.3 mg/g compared to LC × LC's 1.6 mg/g—an approximate 4-fold inflation. Similarly, C at the RL stage showed 1D-LC values of 2.5 mg/g versus LC × LC's 1.2 mg/g, indicating a 108% overestimation. EGCG, the most abundant catechin, demonstrated consistent overestimation patterns, with 1D-LC measuring approximately 27.9 mg/g in WT compared to LC × LC's 23.8 mg/g. In contrast, the two theaflavin compounds revealed an opposite trend where 1D-LC systematically underestimated concentrations relative to LC × LC. For TFDG at the OX stage, LC × LC measured 2.2 mg/g compared to 1D-LC's 1.1 mg/g—a 2-fold difference. This discrepancy intensified at the DR stage, where 1D-LC yielded 1.2 mg/g versus LC × LC's 3.3 mg/g, indicating that 1D-LC underestimated the TFDG content by approximately 64%. TF exhibited a similar pattern, though less pronounced, with 1D-LC measuring 3.4 mg/g at the DR stage compared to LC × LC's 3.9 mg/g (11% underestimation). This bidirectional bias—overestimation for catechins and underestimation for theaflavins—suggests that 1D-LC suffers from complex, compound-specific matrix interferences. By integrating quantitative results (Fig. 3) with co-elution profiles (Fig. 1, Table 1) and matrix effect data (Fig. 2), the mechanistic basis for bidirectional quantification bias becomes clear as discussed below: Catechin overestimation arises from: (1) co-elution with ion-enhancing polar compounds concentrated during withering, reflected in ME = 140–200% (Fig. 2); (2) relatively stable feature density in early-to-mid retention regions (Fig. 1), dominated by small-molecule compounds that enhance rather than suppress ionization; and (3) early elution (RT 12.000–34.000 min) avoiding the most suppressive late-eluting polymers. The net result is 20–300% inflation during WT/RL, normalizing during OX/DR as enhancing species are consumed. Theaflavin underestimation arises from: (1) co-elution with ion-suppressing high-MW polymers (thearubigins, quinone adducts) that sequester ESI charges, reflected in ME = 65–80% (Fig. 2); (2) explosive feature accumulation in late-elution regions (4.7-fold increase, Fig. 1), creating severe charge competition; and (3) late elution (RT 42.000–43.000 min) placing theaflavins within the most congested zone containing >60 co-eluting suppressors (Table 1, R6). The net result: 30–65% signal loss relative to true concentrations. LC × LC corrects both biases by separating catechins from ion-enhancing co-eluters (normalizing ME to 95–110%, Fig. 2) and isolating theaflavins from ion-suppressing polymers (normalizing ME to 95–125%). This provides quantitative data accurately reflecting the underlying biochemistry and chemistry of black tea processing. While Fig. 3 illustrates the quantitative differences in concentration, Table 3 systematically summarizes the accuracy of the ‘change trends’ (i.e., increase, decrease, or stable) reflected by each method. As detailed in Table 3, the two methods frequently reported contradictory trends. For instance, catechin (C) appeared to increase during WT and RL in 1D-LC (↑) but remained stable in LC × LC (→), reflecting a false accumulation trend caused by matrix effects. Similarly, epicatechin (EC) appears stable during WT in 1D-LC (→) but actually decreases according to LC × LC (↓), masking early oxidative losses. Most critically, theaflavins (TF, TFDG) appear to plateau during DR in 1D-LC (→) but continue to accumulate in LC × LC (↑), indicating that 1D-LC fails to capture the full extent of late-stage oxidation. Of the 10 compounds examined, only 2 (EGC and EGCG) showed consistent trends between methods (√), while 9 compounds exhibited trend inconsistency (×). This comparison highlights that 1D-LC fails not only in quantification accuracy but also in reflecting the correct direction of compound evolution during tea processing due to matrix effect. Table 3Comparison of 1D-LC and LC × LC for accurately capturing concentration trends of catechins and theaflavins across processing stages.Compound1D-LCLC × LCTrend consistencyWTRLOXDRWTRLOXDREpigallocatechin (EGC)↑→↓↓↑↑↓↓×Catechin (C)↑↑↓↓→→↓↓×Gallocatechin (GC)↑↓↓↓↑→↓↓×Epigallocatechin gallate (EGCG)↓↓↓↓↓↓↓↓√Epicatechin (EC)→↓↓↓↓↓↓↓×Epicatechingallate (ECG)→→↓↓↓↑↓↓×Theaflavin (TF)→→↑→→→↑↑×Theaflavin-3,3′-digallate (TFDG)→→↑→→→↑↑×Theaflavin-3-gallate (TF3G)––––→→↑↑×Theaflavin-3′-gallate (TF3’G)––––→→↑↑ד↑ / ↓ / →” indicate an increase, decrease, or no significant change in compound concentration, respectively.“-” indicate that compound was not detected or successfully separated in the 1D-LC method.“√ / ×” indicate whether the trend observed in 1D-LC is consistent (√) or inconsistent (×) with the trend observed in LC × LC. Comparison of 1D-LC and LC × LC for accurately capturing concentration trends of catechins and theaflavins across processing stages. “↑ / ↓ / →” indicate an increase, decrease, or no significant change in compound concentration, respectively. “-” indicate that compound was not detected or successfully separated in the 1D-LC method. “√ / ×” indicate whether the trend observed in 1D-LC is consistent (√) or inconsistent (×) with the trend observed in LC × LC. These quantitative data unequivocally demonstrate that LC × LC provides superior accuracy by minimizing co-elution interferences through better separation, particularly crucial for oxidation products such as theaflavins. The systematic overestimation and underestimation by 1D-LC likely stems from MEs in complex black tea matrices. The profound quantification discrepancies between 1D-LC and LC × LC have direct consequences. The bidirectional quantification biases observed in 1D-LC have direct and significant consequences for tea quality assessment and commercial grading systems. Premium black teas are typically characterized by a high theaflavin content (particularly TFDG), which contributes to the desirable bright, brisk, and astringent characteristics valued in high-grade products (Jiang et al., 2025; Qiu et al., 2024). Such limitations can lead to substantial misinterpretation of chemical transformations during processing; for example, a 64% underestimation of theaflavins would obscure the true extent of oxidation and distort our understanding of how key quality-related metabolites evolve. For example, a black tea with a true TFDG content of 3.0 mg/g (premium grade) would be measured at only 1.8 mg/g by 1D-LC, potentially relegating it to a lower commercial grade with corresponding price penalties. Conversely, the 2–4-fold overestimation of residual catechins during early black tea processing stages could mask incomplete oxidation, leading to acceptance of under-fermented batches that should be rejected for insufficient colour and flavor development. This comparison underscores the critical importance of employing LC × LC for reliable multi-component quantification in complex natural product analyses. Conclusion This study systematically evaluated matrix-induced quantification bias during black tea processing using 1D-LC and LC × LC coupled with QTOF. Profiling across four processing stages revealed extensive uncharacterized matrix complexity (3617 ion features, 74 identified), with co-elution evolving markedly during oxidation and drying. In 1D-LC, matrix effects caused bidirectional quantification bias—catechins were generally overestimated and theaflavins generally underestimated—due to co-elution with highly abundant polymeric oxidation products. It should be emphasized that the overestimation observed in this study represents a worst-case scenario for a specific catechin (GC) at a specific stage (withering), and does not imply that all catechins are equally overestimated across all stages. For most catechins and stages, the overestimation by 1D-LC was 20–30%, and relative trends between 1D-LC and LC × LC remained largely consistent (e.g., decrease of EGCG, increase of TFDG). Therefore, our conclusion does not invalidate previous 1D-LC studies but rather refines our understanding of matrix-induced quantification bias. In this case, 1D-LC may distort the apparent kinetics of catechin depletion and theaflavin formation and misrepresented key quality-related markers. In contrast, LC × LC's enhanced separation successfully resolved co-eluting interferents, minimized matrix effects across all stages (ME = 0.75–1.25), and enabled accurate quantification of previously unresolvable positional isomers (TF3G and TF3’G). This allowed calculation of the monogallate/digallate theaflavin ratio (0.52–0.54), a critical quality marker for tea brightness and astringency. The LC × LC method demonstrated excellent analytical performance (R2 ≥ 0.9976, RSD ≤ 5.71%) while maintaining accuracy across dramatically different matrix compositions from fresh leaves to finished tea, meeting regulatory requirements for method robustness. These improvements enable a more accurate understanding of catechin oxidation and theaflavin formation pathways during black tea processing. This has important implications for tea chemistry research, as it allows researchers to reliably monitor metabolite transformations and investigate reaction mechanisms in complex matrices. Despite the demonstrated advantages, this study has certain limitations. The investigation was restricted to a single cultivar of Chinese black tea (Camellia sinensis var. sinensis) from one geographical region. Given that matrix composition can vary by cultivar and processing style (e.g., CTC vs. Orthodox), future research should validate this LC × LC strategy across a broader range of tea varieties. Additionally, while comprehensive 2D-LC offers superior resolution, the requirement for specialized instrumentation and complex data processing may currently limit its widespread adoption in routine quality control laboratories compared to conventional 1D-LC systems. Notwithstanding these constraints, the results of this study demonstrate that LC × LC provides a superior and reliable solution for better quantification in matrix-evolving food systems without extensive sample cleanup or isotope-labeled standards. The analytical framework presented here—combining profiling to identify co-elution patterns, matrix effect evaluation to quantify biases, and LC × LC separation to resolve interferents—is promising for applying to other foods undergoing fermentation or oxidative processing. CRediT authorship contribution statement Shanbo Zhang: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Resources, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Kim-Huey Ee: Writing – review & editing, Supervision, Project administration, Methodology, Investigation, Formal analysis, Conceptualization. Yunle Huang: Writing – review & editing, Methodology, Investigation. Rui Min Vivian Goh: Writing – review & editing, Methodology, Investigation. Lingyi Li: Resources. Lionel Jublot: Writing – review & editing, Supervision, Resources, Project administration, Funding acquisition. Chee Sian Gan: Writing – review & editing, Supervision, Resources, Project administration. Shao Quan Liu: Writing – review & editing, Supervision, Resources, Project administration, Funding acquisition. Bin Yu: Writing – review & editing, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization. Ethical statement Ethics approval was not required for this research. Funding This research was partially funded by Mane SEA PTE LTD. 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. Acknowledgements The authors were grateful to Mane SEA PTE LTD for providing technical assistance and funding during the study. The authors would also like to thank Mr. Teck Maan Lim, Dr. Eric Tan, and Dr. Eugene Goh from Agilent Technologies Singapore (Sales) PTE LTD. for their valuable technical assistance. Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.fochx.2026.103991. Contributor Information Shao Quan Liu, Email: fstlsq@nus.edu.sg. Bin Yu, Email: gsgpbiy@gmail.com. Supplementary data Supplementary material Supplementary data about the details of tea processing parameters, putative identified compounds in Chinese black tea, and chromatograms of demonstrated complex Chinese black tea matrix. Data availability The data that support the findings of this study are available upon request. References Associated Data Supplementary Materials Supplementary material Supplementary data about the details of tea processing parameters, putative identified compounds in Chinese black tea, and chromatograms of demonstrated complex Chinese black tea matrix. Data Availability Statement The data that support the findings of this study are available upon request.