Then, an UPLC-ESI-MS/MS system (UPLC, ExionLC™ AD, https://sciex.com.cn/) and Tandem mass spectrometry system (https://sciex.com.cn/) were used to analyze the sample extracts. Each parameter choice was carefully considered based on established scientific principles and previous research in the field (Li et al., 2022; Cao et al., 2023). The UPLC analytical conditions were set as follows: (1) Column: Agilent SB-C18 (1.8 μm, 2.1 mm * 100 mm); (2) Mobile phase: Solvent A, pure water with 0.1% formic acid; Solvent B, acetonitrile with 0.1% formic acid; (3) Gradient elution: The analysis began with 95% A and 5% B. Over 9 min, a linear gradient was applied to reach 5% A and 95% B, which was maintained for 1 min. Then, the composition was adjusted to 95% A and 5.0% B over 1.1 min and held for 2.9 min; (4) Flow velocity, 0.35 mL/min; column oven, 40°C; injection volume, 2 μL. The effluent was alternately connected to an ESI-triple quadrupole-linear ion trap (Q-TRAP)-MS.

The ESI source operation parameters were set as follows: Source temperature, 500°C; Ion spray voltage (IS), 5500 V (positive ion mode)/-4500 V (negative ion mode); Ion source gas I (GSI), II (GSII), and curtain gas (CUR) were set at 50, 60, and 25 psi, respectively; Collision-activated dissociation (CAD) was set to high. QQQ scans were acquired as multiple reaction monitoring (MRM) experiments with collision gas (nitrogen) set to medium. Declustering potential (DP) and collision energy (CE) for individual MRM transitions were optimized as needed. Specific MRM transitions were monitored based on the metabolites eluted during each period.

For qualitative analysis, the secondary mass spectrum data from UPLC-MS/MS were acquired and pre-processed using Analyst software (Version 1.6.3). To ensure the accurate metabolite annotation, isotopic signals, duplicate signals containing K⁺, Na⁺, and NH₄ ⁺ ions, and repeated fragment ions from other larger molecular weight substances were excluded from the analysis. The data were then annotated using the self-built Metware database (MWDB, Metware Biotechnology Co., Ltd., Wuhan, China).

For quantitative analysis, the MRM mode of triple-quadrupole mass spectrometry was employed. Peak detection, integration and correction were performed using MultiaQuant™ software. The corrected peak area for each chromatographic peak was used as an indicator of the relative concentration of the metabolites.

All metabolites of E. japonica identified through widely targeted metabolomics were queried in the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP, https://old.tcmsp-e.com/tcmsp.php). Metabolites meeting the criteria of oral bioavailability (OB) ≥ 5% and drug-likeness (DL) ≥ 0.14 (Li et al., 2022) were considered key bioactive ingredients in Traditional Chinese Medicines derived from E. japonica. The related-targets and related-diseases information of these identified metabolites were listed as the TCMSP database annotation.

Based on the disease-related annotation in the TCMSP database, the active pharmaceutical ingredients in E. japonica with function of anti-cancer/tumor, anti-Alzheimer’s disease, analgesics, anti-myocardial infarction, anti-inflammatory, anti-arthritis, anti-gestational hypertension, anti-stroke, and anti-cardiovascular disease were summarized. The identification process of these active pharmaceutical ingredients followed previous methodologies with some modifications (Li et al., 2022; Xia et al., 2023).

To identify differential metabolites among different E. japonica tissues, the PCA and HCA were performed. The PCA results indicated clear separation among tissue groups. High aggregation of QC samples confirming the stability and reliability of the chromatography and quality detection system. As shown in Figure 4A , the two principal components (PC1 and PC2) explained 45.74% of the cumulative variance (PC1: 24.67%, PC2: 21.07%). Notably, PC1 and PC2 primarily differentiated the Inf and LR from the other four tissues. The remaining tissues, MR, S, L, and Fr were relatively clustered but still distinguishable along PC1 and PC2. These results suggest that the metabolites in E. japonica tissues differ significantly, with Inf and LR showing the most pronounced differences.

In line with the PCA results, HCA also demonstrated significant tissue-specific metabolite differences ( Figure 4B ). In PCA, the four biological replicates of each tissue clustered tightly, and the correlation analysis showed that the Pearson correlation coefficient |r| was close to 1 for all inter-group comparisons ( Supplementary Figure S4 ), confirming high data reproducibility.

OPLS-DA was used to screen differential metabolites among tissues. The OPLS-DA models exhibited high predictability (Q² ≥ 0.991) and strong goodness of fit (R²X ≥ 0.722, R²Y = 1) across all fifteen pairwise comparisons among the six tissues, indicating model stability and reliability ( Supplementary Figure S6 ).

The OPLS-DA model score plots ( Supplementary Figure S7 ) revealed significant separations among tissue groups, suggesting distinct metabolite profiles. The OPLS-DA S-plots, which graphically project specific metabolites, showed that red and green dots represent metabolites with VIP values ≥ 1 and < 1, respectively ( Supplementary Figure S8 ). The top 20 metabolites with higher VIP values for each comparison were shown in Supplementary Figure S9 . Notably, in LR comparisons (except LR vs. Inf), at least 15 of the top 20 metabolites were upregulated in LR compared to other tissues.

Fold change (FC) values were calculated to provide a clearer and more intuitive representation of overall metabolic differences. A dynamic distribution map of metabolite content differences was created, highlighting the top 10 up- and down-regulated metabolites ( Supplementary Figure S10 ). Each pairwise comparison identified more than 1,100 metabolites meeting the criteria of |Log2FC| ≥ 1.0, with the most (1,543) in LR vs. L and the fewest (1,127) in MR vs. S ( Supplementary Table S7 ). Moreover, among the 15 comparisons, LR vs. MR/Fr/Inf/S/L exhibited a higher number of up-regulated metabolites, with 1,056, 974, 825, 843, and 885 metabolites, respectively ( Supplementary Table S7 ).

To further explore tissue differences, DEMs were screened from the 2140 identified metabolites based on FC and VIP scores. Using the criteria (VIP ≥ 1; |Log2FC| ≥ 1.0; p ≤ 0.05), a total of 2045 DEMs were identified in at least one pairwise comparison, accounting for 97.2% of all detected metabolites. As shown in Figures 4C , Supplementary Figure S11 and Supplementary Tables S7 , S8 , over 40% of detected metabolites in all pairwise comparisons were DEMs, with the most (1,335) in MR vs. Inf and the fewest (873) in S vs. L. These results strongly suggest significant variations in metabolite composition and regulation among tissues.

To evaluate metabolite variations across tissues of E. japonica, we performed unit variance scaling on 2045 DEMs and clustered them into 10 subclasses using K-means analysis ( Figure 5 ; Supplementary Table S7 ). This classification revealed tissue-specific metabolite accumulation patterns. For instance, subclass 1 to 6 exhibited high metabolite accumulation in MR, LR, Fr, Inf, S and L, respectively, with 145, 424, 254, 418,76 and 262 DEMs each ( Figure 5 ; Table 1 ). Saccharides, though a minor fraction of all DEMs, were prominent in MR-specific subclass 1( Figure 5A ). Flavonoids dominated subclasses 2 to 6, indicating their prevalence in LR, Fr, Inf, S, and L ( Figures 5B–F ; Table 1 ). Subclass 2 was distinguished by elevated alkaloids (35) and terpenoids (25), with lipids and phenolic acids ranking second to subclass 4. This indicates that flavonoids, phenolic acids, lipids, alkaloids, and terpenoids are the primary metabolites with high accumulation in the lateral roots of E. japonica. Subclass 3 was rich in flavonoids, amino acids and derivatives, and organic acids. Subclass 4, with 418 Inf-concentrated metabolites, included 104 flavonoids, 83 lipids, 65 phenolic acids, 45 amino acids and derivatives, and 32 organic acids, totally accounting for 78.7%. Subclass 5, with 76 stem-enriched metabolites, highlighted flavonoids, phenolic acids, and alkaloids. Subclass 6, boasting 262 leaf-enriched metabolites, featured flavonoids, phenolic acids, and saccharides. Subclasses 7 to 10 had elevated metabolite levels in tissue pairs: LR & Inf (99), LR & S (113), Inf & L (161), and S & L (93), with flavonoids being predominant. Overall, inflorescences and lateral roots of E. japonica had the highest number of metabolites at elevated levels, followed by leaves, stems, fruits, and main roots.

Flavonoids are the most abundant metabolites in E. japonica, with 638 identified across six tissues. They can be further categorized into nine subclasses: flavones, isoflavones, flavonols, flavanones, chalcones, dihydroisoflavones, flavanols, flavanonols, and others ( Supplementary Figure S2A ). Each tissue of E. japonica identified 82.1% to 87.9% of total flavonoids, with the highest proportion found in LR, followed by L, Inf, S, and MR ( Supplementary Figure S12 ). Similarly, LR also contained the highest number of richest flavonoids (30.4%, 194), followed by L and Inf ( Supplementary Figure S5B ). Comparative analyses revealed more up-regulated flavonoids in LR, L, and Inf than in other tissues, while MR contained more down-regulated ones ( Supplementary Table S8 ). In summary, although the number of detectable flavonoid metabolites varied little among the six tissues, the concentration of flavonoids in LR, L, and Inf was significantly higher than that in Fr, S, and MR, highlighting their potential in E. japonica’s medicinal value.

Flavonoids are well-documented for their significant role in managing chronic diseases by inhibiting oxidative damage and persistent inflammation (Ding et al., 2022). Pharmacological data further underscore the importance of flavonoids in disease resistance ( Supplementary Table S9 ). Among the 87 core key active flavonoids, 68 were effective against nine major diseases, many highly accumulated in LR, Inf, and L ( Supplementary Tables S2 , S9 ). Many of these flavonoids are known for their antioxidant, anti-inflammatory, anti-arteriosclerotic, anti-tumor, and cancer therapeutic effects. For instance, wogonin, a flavone with notable antitumor activity and a potential chemosensitizer in cancer therapy (Huynh et al., 2017, 2020; Zhang et al., 2022), was highly concentrated in LR but low in MR/Inf and absent in Fr/S/L ( Figure 6A ). Calycosin and formononetin, isoflavones with strong cardiovascular protective effects (Chen et al., 2022; Ma et al., 2022; Qian et al., 2024), were most abundant in LR, followed by Fr or S ( Figure 6A ). Other flavonoids, such as taxifolin (Topal et al., 2016), glabridin (Parlar et al., 2020), kaempferol (Devi et al., 2015), wighteone (Wang et al., 2024), catechin, catechin gallate and epicatechin gallate (Musial et al., 2020), which exhibit excellent antioxidant and/or anti-inflammatory effects, were also highly accumulated in LR ( Figure 6A ). Similarly, several core key active flavonoids with the highest accumulation in Inf have also been reported to have diverse physiological and pharmacological effects. These include four flavanones (hesperetin, liquiritin, naringenin, cirsiliol, and didymin), three flavones (isovitexin, diosmetin, and luteolin), two flavonols (morin and galangin), and one flavanonol (engeletin) ( Figure 6B ). Isovitexin is a potential inhibitor of multiple tumor cells, including colon, breast, ovarian, prostate, esophageal, and pancreatic cancer, demonstrating its capability for promoting apoptosis in tumor cells (Yang et al., 2013; Ganesan and Xu, 2017; Ghanbari-Movahed et al., 2023). Engeletin, with anti-inflammatory, antioxidant, and immunomodulatory properties (Zhong et al., 2023), is a promising drug candidate. Flavonoids with higher relative contents in L include acacetin, genkwanin, nobiletin, kumatakenin, 8-prenylkaempferol, icaritin, glycitein, etc. ( Figure 6C ). Acacetin has demonstrated anti-inflammatory, antiperoxidative, and anticancer activities, and is a promising potential drug for treating osteoarthritis (Kim et al., 2014; Chen et al., 2020). Icaritin possesses a wide range of pharmacological effects, including alleviating sexual dysfunction, osteoporosis, cardiovascular diseases (Reyes-Hernández et al., 2024), and showing efficacy against solid tumor, particularly advanced hepatocellular carcinoma (Luo et al., 2024). Among the 11 core key active flavonoids with the highest accumulation in Fr, baicalin and hispidulin ( Figure 6D ) are noted for their anti-diabetic, anti-inflammatory, antioxidant, and anticancer properties (Wang et al., 2020; Miao et al., 2024). The six core key active flavonoids with the highest accumulation in S include three flavanols (cinchonain Ic, cinchonain Id, and garbanzol), two isoflavones (flemiphilippinin C, and glycyroside), and one other flavonoid (kushenol O) ( Supplementary Table S1 ). Taken together, many flavonoids with diverse pharmacological activities like anti-cancer/tumor, anti-inflammatory, etc., are predominantly concentrated in LR, Inf, and L, and those flavonoids can be used as lead compounds for the development of potential therapeutic agents. For example, wogonin with high anti-tumor activity can be structurally modified to develop new anti-cancer drugs (Banik et al., 2022). Calycosin and formononetin with cardiovascular-protective effects can be further studied for drug or health-products development.

A comprehensive analysis identified 280 phenolic acids in E. japonica, with LR exhibiting the highest accumulation, followed by S, MR, L, Inf, and Fr ( Supplementary Figure S3 ). Comparative analyses revealed more up-regulated phenolic acids in LR, S, and Inf than in L, MR, and Fr, indicating higher concentrations in the former ( Supplementary Table S8 , Supplementary Figure S13 ). Among these, 21 were key active ingredients, and 12 were active pharmaceutical ingredients ( Supplementary Table S2 , S9 ). Four of these—grandidentatin, salidroside, syringin, and usnic acid—show resistance to nine major diseases ( Supplementary Table S2 ). Salidroside, for instance, has been shown to reduce oxidative substances such as reactive oxygen species and malondialdehyde, thereby delaying the progression of aging-related diseases (Campisi et al., 2019). Though not meeting the criteria for key active ingredients, gallic and ferulic acids were predicted to resist all nine diseases ( Supplementary Table S2 ). Gallic acid is known for its antioxidant, anti-inflammatory, antineoplastic activities, and therapeutic activities against gastrointestinal, neuropsychological, metabolic, and cardiovascular disorders (Kahkeshani et al., 2019). Ferulic acid is effective in treating diabetes, angina pectoris, heart stroke, coronary heart disease, and cardiovascular diseases (Liu et al., 2003; Guo et al., 2008). Of the six highlighted phenolic acids, except gallic acid (high in LR) and usnic acids (high in L), the other four were most abundant in Inf ( Supplementary Figure S25 ). Given their significant antioxidant properties, the high levels of phenolic acids in E. japonica, especially in Inf, LR, and S, suggest its potential for improving human health.

Over 200 amino acids and derivatives were found in each tissue of E. japonica, representing over 90% of the total 220 ( Supplementary Figure S3 ). Notably, Inf, LR, and Fr had higher concentrations ( Supplementary Table S8 , Supplementary Figure S14 ). None were key active ingredients, but 19 were active pharmaceutical ingredients. Six were predicted to correspond to nine major diseases: L-valine, 5-hydroxy-L-tryptophan, L-tryptophan, L-histidine, 5-hydroxy-DL-tryptophan, and N-methylphenylalanine ( Supplementary Table S2 ). L-valine and L-histidine were more prevalent in MR and Fr, respectively, while 5-hydroxy-DL-tryptophan and N-methylphenylalanine were particularly abundant in Inf ( Supplementary Figure S26 ).

In E. japonica, 205 lipids were detected across tissues, with LR containing the most (197) and Fr the least (168) ( Supplementary Figure S3 ). Comparative analyses revealed more up-regulated lipids in Inf and LR, suggesting higher concentrations in these tissues ( Supplementary Table S8 ; Supplementary Figure S15 ). Out of the total lipids, 25 were key active ingredients, and 16 were active pharmaceutical ingredients ( Supplementary Tables S2 , S9 ). Among these, 13 lipids were predicted to protect against nine major diseases, including nine most highly accumulated in Inf (9-hydroxy-10,12,15-octadecatrienoic acid, ethyl linoleate, linoleic acid, octadeca-9,12,15-trienoic acid, pinolenic acid, punicic acid, α-linolenic acid, and γ-linolenic acid, ricinoleic acid), two in LR (arachidonic acid, methyl linolenate), one in MR (elaidic acid) and one in Fr (stearic acid) ( Supplementary Figure S27 ). Ricinoleic acid, linoleic acid, elaidic acid and stearic acid have also accumulated considerable amounts in other tissues ( Supplementary Figure S27 ). Additionally, arachidonic acid and ethyl linoleate met criteria for potential drug candidates, with arachidonic acid playing a crucial role in cardiometabolic diseases (Hu et al., 2024), indicating its therapeutic potential in cardiovascular health. The higher relative content of these lipids in Inf and LR may enhance their value in protecting human health.

Among the 186 alkaloid metabolites detected in E. japonica, LR had the highest number of alkaloids (163), as well as richest accumulated alkaloids (53, 28.49%) ( Supplementary Figure S5F ). Differential expression and heatmap cluster analysis indicated a higher relative content of alkaloids in LR, Inf and L ( Supplementary Table S8 ; Supplementary Figure S16 ). Nine alkaloids were core key active ingredients: lauroscholtzine, N-cis-feruloyltyramine, tetrahydroprotopapaverine, p-coumaroyltyramine, sophocarpine, sparteine, β-isosparteine, isomatrine, and moupinamide. Notably, the first four were core key active pharmaceutical ingredients, predicted to address nine major diseases, and significantly accumulated in LR ( Supplementary Table S2 ; Supplementary Figure S28 ). Sophocarpine was also suggested to have anti-nociceptive, anti-inflammatory, neuroprotective, anti-tumor, and immune regulatory functions (Zhou et al., 2022), with higher relative content found in Fr and Inf. Additionally, sparteine and β-isosparteine were found to be significantly accumulated in S and were predicted to have anti-Alzheimer’s and analgesic effects, while isomatrine was notably present in Fr, L and Inf, and moupinamide was concentrated in LR.

Of the 121 organic acids identified in E. japonica, the detectable quantities in each tissue ranged from 109 to 118 ( Supplementary Figure S3 ). A high relative abundance of organic acids was also observed in Inf, Fr, and L, while MR and S exhibited lower levels ( Supplementary Table S8 ; Supplementary Figure S17 ). Among the 41 organic acids recognized as components of traditional Chinese medicines in TCMSP, only two—anacardic acid and tianshic acid—were identified as key active ingredients ( Supplementary Table S2 ). Anacardic acid, predicted to possess analgesic and anti-inflammatory properties, was found to be highly concentrated in Fr, whereas tianshic acid was relatively more abundant in Inf ( Supplementary Figure S29 ).

In MR, all lignans and coumarins identified in E. japonica were detected, totaling 103, while Fr detected the fewest, with only 76 ( Supplementary Figure S2 ). Notably, LR contained the largest number of the richest accumulated lignans and coumarins, totaling 33 ( Supplementary Figure S5H ). Moreover, LR, S, and L also contained relatively high levels of lignans and coumarins ( Supplementary Table S8 ; Supplementary Figure S18 ). A total of 20 lignans and coumarins were identified as key active ingredients, among which 10 were further categorized as core key active ingredients with potential for drug development. This subset included six lignans (acanthoside B, diphyllin, fargesin, glycyrin, picraquassioside C, and sesamin) and four coumarins (coumestrol, dalbergin, fraxin, and stevenin) ( Supplementary Table S2 ). Furthermore, diphyllin, glycyrin, sesamin, dalbergin, and stevenin were predicted to be effective against nine major diseases, while coumestrol and fraxin were found to confer resistance to eight major diseases, excluding cardiovascular disease ( Supplementary Table S2 ). The pharmacological properties of these lignans and coumarins have been continuously validated; for example, sesamin is known for its antioxidative, anti-cancerogenic, anti-inflammatory, anti-proliferative, anti-hypertensive, anti-melanogenesis, and promoting effect in treating osteoporosis (Yang et al., 2022; Dossou et al., 2023), highlighting its potential for various therapeutic applications. Coumestrol has also demonstrated anti-tumor activity by facilitating apoptosis in colorectal cancer cells (Geng et al., 2024). Here, considerable accumulation of sesamin was detected in all six tissues of E. japonica. Additionally, fargesin, picraquassioside C, acanthoside B, and coumestrol exhibited relatively high levels in MR, LR, and S. Fraxin and dalbergin were notably concentrated in LR, while stevenin was found in both LR and Fr, and glycyrin and diphyllin were primarily present in S and L, respectively ( Supplementary Figure S30 ). These findings suggest that lignans and coumarins in various tissues of E. japonica significantly contribute to its pharmacological efficacy.

A total of 72 nucleotides and derivatives were detected in E. japonica, with detectable quantities in each tissue ranging from 63 to 71 ( Supplementary Figure S3 ). Inf and LR had higher levels of these compounds ( Supplementary Table S8 ; Supplementary Figure S19 ). Six were identified as key active ingredients: 2’-deoxyadenosine, adenosine, cordycepin, guanosine, uridine 5’-diphosphate, and uridine 5’-monophosphate ( Supplementary Tables S2 , S9 ). Their concentrations across various tissues were presented in Supplementary Figure S31 . Additionally, 2’-deoxyadenosine and cordycepin were predicted to be associated with nine major diseases ( Supplementary Table S2 ). Existing studies have also confirmed that cordycepin, recognized as a potential anti-cancer/tumor agent, exhibits complementary therapeutic activities in promoting apoptosis, anti-proliferation, and anti-metastasis in cancer cells (Yoon et al., 2018; Schwenzer et al., 2021).

The highest number of terpenoids was detected in LR (67), followed by MR (65) ( Supplementary Figure S3 ). Differential metabolite analysis revealed that the relative contents of many terpenoids in LR were significantly higher than those in other tissues ( Supplementary Table S8 ; Supplementary Figure S20 ). A total of 21 terpenoids were identified as key active ingredients, comprising 16 triterpenes, four diterpenoids, and one triterpene saponin ( Supplementary Table S9 ). Among these, two triterpenes (asiatic acid, and mangiferonic acid) and three diterpenoids (gibberellin A3, isopimaric acid, and kaurenoic acid) met the criteria of core key active ingredients ( Supplementary Table S2 ; Supplementary Figure S32 ). Asiatic acid, despite no TCMSP disease annotations, has shown anti-inflammatory, antioxidant, anti-fibrosis, and anti-tumor properties in diverse disease models (Lv et al., 2017; Chung et al., 2021), and was most abundant in LR, indicating the tissue’s potential health benefits.

A total of 22 quinones were detected in E. japonica, with approximately 21 detected in all tissues except for Inf, with only 13 identified in Inf ( Supplementary Figure S2 ). LR and Fr contained more up-regulated quinones than down-regulated ones compared to other tissues, indicating higher overall quinone levels in LR and Fr ( Supplementary Table S8 ; Supplementary Figure S21 ). Seven quinones were identified as key active ingredients, including three highly accumulated quinones in LR (1,2,4-trihydroxyanthraquinone, physcion, and rheic acid), two highly accumulated quinones in Fr (emodin-8-O-glucoside, and helminthosporin), one quinone shared between LR and Fr (rubiadin-1-methyl ether), and one quinone (embelin) that was commonly found in all tissues ( Supplementary Table S2 ; Supplementary Figure S33 ). Furthermore, embelin, emodin-8-O-glucoside, and rheic acid were classified as core key active ingredients. Additionally, 1,2,4-trihydroxyanthraquinone, physcion, rheic acid, and embelin were annotated as active pharmaceutical ingredients associated with nine or eight major diseases ( Supplementary Table S2 ). The higher accumulation of these quinones in LR and/or Fr suggested that they significantly contribute to the health benefits attributed to these tissues.

Tannins were predominantly detected in LR, where 14 out of 15 tannins were identified, with 10 exhibiting the highest relative content or accumulating exclusively in LR ( Supplementary Figures S3 , S5L , S22 ). Among the 15 tannins, only seven were annotated in TCMSP, of which four were annotated as key active ingredients: 3,3’-di-O-methylellagic acid 4’-glucoside, ellagic acid, procyanidin B1, and procyanidin B4 ( Supplementary Tables S2 , S9 ). Metabolite content analysis revealed that ellagic acid accumulated at comparable levels across all six tissues, while 3,3’-di-O-methylellagic acid 4’-glucoside mainly accumulated in high concentrations in S and L. In contrast, procyanidin B1 and procyanidin B4 were highly or exclusively accumulated in LR ( Supplementary Figure S34 ).

In addition to the aforementioned metabolites, a total of 207 metabolites from eight other categories were detected in E. japonica ( Figure 2B ), with quantities ranging from 191 to 200 in each tissue ( Supplementary Figure S3 ). Besides saccharides being most abundant in MR, other types of metabolites were also mainly accumulated in LR and Inf ( Supplementary Table S8 ; Supplementary Figures S23 , S24 ). Of the 64 metabolites annotated in TCMSP, only 12 were annotated as key active ingredients, while 25 were identified as active pharmaceutical ingredients, with eight metabolites co-annotated ( Supplementary Table S2 ). The 12 key active ingredients included three chromones, three saccharides, three vitamins, two ketone compounds, and 3’-O-methylorobol ( Supplementary Table S2 ; Supplementary Figure S35 ). Among these metabolites, capillarisin, vitamin K1, and 3’-O-methylorobol met the criteria of OB ≥ 30% and DL ≥ 0.18, and were associated with nine or eight major diseases, indicating their potential for new drug development ( Supplementary Table S2 ).

In summary, E. japonica’s tissues showed distinct metabolite profiles. LR had higher accumulations of most metabolite classes, especially flavonoids, phenolic acids, alkaloids, lignans and coumarins, terpenoids, quinones, and tannins. Inf was rich in many metabolite classes except lignans and coumarins, quinones, and tannins. In other tissues, the accumulation of certain metabolites was relatively abundant. For instance, the content of flavonoids, alkaloids and organic acids was relatively high in L; phenolic acids, lignans and coumarins were richly accumulated in S; organic acids and quinones were also abundant in Fr. However, except for saccharides that predominantly accumulated in MR, the content of other metabolites in MR was comparatively lower than that in other tissues. These findings indicate that for E. japonica, with its roots used as folk medicine, the medicinal efficacy likely derives mainly from LR, as LR generally contains a higher content of pharmacologically active components than MR.

Overall, the diverse chemical composition across E. japonica tissues points to its significant pharmacological potential. Notably, LR and Inf demonstrate the greatest application potential due to their rich profiles of bioactive compounds, suggesting they could serve as key sources for developing treatments for various diseases, particularly for treating chronic and multifactorial diseases. This study underscores the pharmacological promise of E. japonica in medicinal applications, warranting further investigation into these specific tissues for drug development.

KEGG pathway enrichment analysis was conducted based on the characteristics of the DEMs. In KEGG enrichment analysis, the rich factor reflects the degree of enrichment; a value closer to 1 indicates a higher proportion of annotated metabolites within a corresponding pathway that are classified as DEMs. The top 20 pathways identified via KEGG enrichment analysis were shown in Supplementary Table S10 , Figure 7 and Supplementary Figures S37–39 .

Among the top 20 pathways across the 15 comparisons, a total of 88 pathways were identified, with 32 pathways significantly enriched (p < 0.05) in at least one comparison ( Supplementary Table S10 ; Supplementary Figure S37 ). Notably, six pathways—biosynthesis of isoflavones aglycones I, II, and III, and biosynthesis of flavones aglycones I, II, and III—ranked among the top 20 pathways in most comparisons and were significantly enriched in several cases. Additionally, ten pathways, including starch and sucrose metabolism, flavonoid biosynthesis, isoflavonoid biosynthesis, galactose metabolism, pentose and glucuronate interconversions, biosynthesis of cofactors, ascorbate and aldarate metabolism, biosynthesis of kaempferol aglycones I, carbon fixation in photosynthetic organisms, and nucleotide metabolism, were significantly enriched in at least two comparisons ( Supplementary Table S10 ; Supplementary Figure S37 ). Overall, the pathways mentioned above are likely responsible for the significant differences in chemical composition observed among the six tissues of E. japonica.

The DA score is a pivotal metric for metabolite changes within metabolic pathways, with values closer to 1 indicating more up-regulated DEMs and -1 indicating more down-regulated DEMs. The top 20 pathways, ranked by p-values, were detailed in Supplementary Table S10 , yielding 300 records across 88 pathways from 15 comparisons ( Supplementary Figure S37 ). Notably, the “biosynthesis of flavones aglycones I” and “biosynthesis of isoflavones aglycones I” pathways were most prominent, featuring in 12 of the 15 comparisons. Other notable pathways included “biosynthesis of isoflavones aglycones II,” “biosynthesis of flavones aglycones II,” “biosynthesis of flavones aglycones III,” “isoflavonoid biosynthesis,” and “flavonoid biosynthesis”, which appeared in at least eight comparisons. This underscores the significant divergence in flavonoid and isoflavonoid biosynthesis across tissues. Up-regulated DEM distribution revealed that the expression of “flavonoid biosynthesis” pathway was most active in Inf, while “isoflavonoid biosynthesis” was most abundant in LR ( Figure 8 ). In similar studies of Astragalus membranaceus and Sophora flavescens, flavonoids have also been confirmed as their main bioactive components, and these components are highly accumulated in the roots. The studies have also identified flavonoid and isoflavonoid biosynthesis as critical pathways, highlighting their conserved functions in medicinal legumes (Wu et al., 2020; Wei et al., 2021). Revealing flavonoid and isoflavonoid biosynthetic pathways and their regulation in medicinal legumes will provide a key theoretical basis for cultivating high-bioactivity plants in the future.

In the comparative analysis of each tissue against others, the top 20 pathways were selected, and DA scores for 100 pathways per tissue were statistically analyzed. Inf vs. MR/LR/Fr/S/L showed the most pathways (77) with DA scores above zero, indicating higher expression, followed by LR (52), Fr (46), L (42), S (34), and MR (21) ( Figure 8C ). These findings suggest that the expression of numerous pathways in Inf and LR was more abundant than in other tissues, whereas MR showed relatively fewer pathways expressed at higher levels compared to the other tissues.

In the MR vs. LR/Fr/Inf/S/L comparisons, only 6, 3, 2, 4, and 6 pathways exhibited DA scores greater than zero, respectively ( Figures 7D, E ; Supplementary Figures S39A–C ). Additionally, only two pathways—”starch and sucrose metabolism” and “galactose metabolism”—in the MR vs. L, were significantly enriched with DA scores above zero, indicating relatively high expression of these pathways in MR. In contrast, in the LR vs. MR/Fr/Inf/S/L comparisons, more pathways had DA scores greater than zero ( Figures 7D, F ; Supplementary Figures S39D–F ). Besides those involved in isoflavonoid biosynthesis, metabolites related to “arachidonic acid metabolism,” “biosynthesis of coumarins I,” “biosynthesis of various alkaloids,” “alpha-linolenic acid metabolism,” and “benzoxazinoid biosynthesis” were also relatively abundant in LR. Inf vs. MR/LR/Fr/S/L comparisons highlighted a greater number of pathways with higher DA scores ( Figures 7B, C ; Supplementary Figures S39G, J, K ), indicating that many pathways were expressed more abundantly in Inf than in other tissues. In addition to pathways associated with flavonoid biosynthesis, pathways with high expression in Inf also included “butanoate metabolism,” “pyruvate metabolism,” “apigenin O-glycosides biosynthesis,” “apigenin C-glycosides biosynthesis,” “alanine, aspartate and glutamate metabolism,” and “biosynthesis of kaempferol aglycones I”. In the comparisons of Fr vs. MR/LR/Inf/S/L, S vs. MR/LR/Fr/Inf/L, and L vs. MR/LR/Fr/Inf/L, the number of pathways with high DA scores varied ( Supplementary Figure S39 ). The expression levels of several pathways in these tissues were also noteworthy. For example, the expression of “luteolin aglycones biosynthesis” in Fr was more abundant than in MR/LR/S, while “pentose and glucuronate interconversions” was expressed at higher levels in L compared to MR/LR/Inf/S. Conversely, the expression of “starch and sucrose metabolism” in L was significantly lower than in MR/Fr/Inf/S.