An integrated metabolomics and network pharmacology approach to reveal the antioxidant components and mechanisms of wild jujube (Ziziphus jujuba var. spinosa) fruits across China

Abstract

Wild jujube serves as an important source of natural antioxidants and holds significant economic value in the food and health industries. However, the core antioxidant components in its fruits and their mechanisms of action remain unclear, and the substantial variation in metabolite composition across different provenances severely hinders the development of functional wild jujube products. In this study, untargeted metabolomics combined with network pharmacology was employed to screen 87 potentially active components from the metabolic profile of wild jujube and to identify 41 core antioxidant targets. Among these, seven targets—TP53, AKT1, SRC, STAT3, JUN, EP300, and ESR1—were strongly correlated with antioxidant activity. On the basis of topological and Pearson correlation analyses, 26 key antioxidant compounds were screened from the metabolic profile of wild jujube. Finally, molecular docking revealed the most stable binding pairs: cymarin–ESR1 (−11.3 kcal/mol), procyanidin B1–SRC (−10.8 kcal/mol), and Licoisoflavone A–JUN (−9.9 kcal/mol). This study systematically elucidates the metabolic characteristics of wild jujube from different provenances, provides an in-depth investigation of its antioxidant active ingredients and their mechanisms of action, reveals the physiological functions of wild jujube, and establishes a theoretical foundation for the extraction of its bioactive compounds and the development of functional health foods.

Highlights

• •Wild jujube pulp demonstrated high antioxidant activity.
• •2895 metabolites were identified in pulp samples from six provinces, including 87 potential active compounds.
• •Network pharmacology identified seven core antioxidant-related targets: TP53, AKT1, SRC, STAT3, JUN, EP300, and ESR1.
• •The most favorable binding affinities occurred in cymarin-ESR1, procyanidin B1-SRC, and Licoisoflavone A-JUN complexes.

1. Introduction

Oxidative stress occurs when the levels of reactive oxygen species (ROS) exceed the body's ability to eliminate them, leading to damage to key cellular components (Inupakutika, Sengupta, Devireddy, Rajeev and Azad&Ron Mittler., 2016, Sies and Carsten Berndt&Dean P. Jones., 2017). Recent studies have shown that factors such as pollution, radiation, and poor dietary habits can cause excessive ROS accumulation, thereby inducing oxidative stress and further promoting the development of chronic diseases, including neurodegenerative diseases, diabetes, and cancer (Nam et al., 2024). The body maintains this balance through a combination of endogenous and exogenous antioxidant substances, making exogenous antioxidants essential for sustaining redox regulation (Tumilaar et al., 2024). However, synthetically produced exogenous antioxidants may trigger a range of adverse effects, such as cytotoxicity (Han et al., 2021; Oubannin et al., 2024; Wang et al., 2025). In contrast, natural antioxidants found in fruits—such as phenolic, flavonoid, and terpenoid compounds—are safe and non-toxic(Akbari et al., 2022). These substances can directly scavenge free radicals and modulate upstream factors in antioxidant signaling pathways, thereby alleviating oxidative stress-induced cellular damage.

Wild jujube (Ziziphus jujuba var. spinosa), a medicinal and edible wild plant, is an important source of natural antioxidants. Its seeds are rich in high-quality antioxidants such as flavonoids, phenolics, vitamin C, and terpenoids (Li et al., 2024). However, the fruit of wild jujube is often treated as a byproduct during seed processing and remains underutilized, leading to resource waste and increased environmental pressure (Guo et al., 2019). Recent studies have shown that wild jujube fruit also contains natural antioxidant components, including flavonoids, total phenolics, and polysaccharides, making it a valuable source of natural antioxidants (Wang et al., 2022). Many researchers have confirmed the strong antioxidant activity of triterpenoid saponins from wild jujube fruit through in vitro experiments and have evaluated their total flavonoid and total phenolic contents along with their antioxidant capacities (Sun et al., 2012). These findings indicate that wild jujube fruit, as a plant-derived natural antioxidant, can not only meet the market demand for healthy food products and promote innovation in the food processing industry but can also create new opportunities for the application of agricultural products in fields such as food, medicine, and agriculture. However, as a wild resource, jujube from different provenances shows considerable variation in metabolite composition, resulting in inconsistent raw material quality during processing and posing significant challenges for deep processing of jujube fruit.

In recent years, multiomics technologies and computational simulation methods have been widely applied in food science. The integration of metabolomics with network pharmacology and molecular docking provides a more comprehensive strategy for exploring the complex relationships between food and human health. Although originally developed for drug discovery, this approach has been increasingly adopted to investigate functional components in foods(Duan et al., 2024). For example, the antibacterial activity of honey, an important medicinal and edible food, significantly differs among different types of honey. The combination of metabolomics, network pharmacology, and molecular docking has clarified not only the variations in metabolic composition among different honeys but also the antibacterial metabolites and their targets. These results revealed new mechanisms through which honey regulates oxidative stress and inhibits bacterial growth, providing a solid foundation for the further development of honey in the food industry(Yu et al., 2024). Using a similar approach, active ingredients for treating type 2 diabetes have been screened from Rosa roxburghii fruit, and their mechanisms of action have been preliminarily elucidated, laying a theoretical foundation for further research on its medicinal components (Shen et al., 2023). This methodology has also been successfully applied in the development of other food products, such as pomegranate juice (Smaoui et al., 2019) and strawberries (Zeng, et al., 2025). Thus, the application of this approach can promote the diversified development of medicinal and edible plants in both the food and pharmaceutical sectors. Currently, systematic analyses of the core antioxidant components in jujube fruit metabolites and their mechanisms of action are rare, hindering their development in the food industry and pharmaceutical field. This represents a major limitation in the high-value processing of wild jujube byproducts.

In this study, samples were collected from six different provinces to ensure broad coverage of the metabolite composition of wild jujube. Untargeted metabolomics was employed to analyze differences in metabolic profiles among wild jujubes from various provinces. Combined with network pharmacology and molecular docking, this study aims to reveal key antioxidant compounds in wild jujube and their potential mechanisms of action. This research seeks to promote a comprehensive understanding of wild jujube metabolites, identify core antioxidant components, provide new insights into how wild jujube metabolites inhibit oxidative damage and promote human health, and offer a theoretical basis for advancing the development of wild jujube fruit byproducts in the food industry and pharmaceutical fields.

2. Materials and methods

2.1. Materials

In this study, wild jujube fruits from six distinct provinces preserved at the National Germplasm Repository of Wild Jujube, Shenyang Agricultural University, were used. The specific sampling locations are shown in Fig. 1, and the correspondence between the sampling site codes and province information is detailed in Supplementary Table 1. Upon fruit maturation, sixty fruits were randomly harvested from five individual plants per province. These samples were immediately flash-frozen in liquid nitrogen and allocated into six biological replicates for untargeted metabolomic profiling and physiological parameter analysis.

Fig. 1.

Sampling Map of Wild Jujube Provenances.

2.2. Antioxidant capacity in vitro

2.2.1. DPPH radical scavenging capacity (DPPH)

The DPPH radical scavenging assay was conducted with modifications based on the methodology summarized by (Molyneux, 2004) The reaction mixtures contained 0.8 mL of diluted wild jujube fruit extract and 1.2 mL of 30 μg/mL DPPH solution. After homogenization, the samples were incubated in the dark at 25 °C for 30 min. The absorbance was measured at 515 nm. In the control assays, methanol was substituted for the fruit extract under identical conditions. The DPPH radical scavenging rate was calculated according to Eq. (1). A standard curve was generated by regressing the DPPH radical scavenging rates against the Trolox concentrations (50–200 μg/mL). All the assays were performed in sextuplicates. The results are expressed as TEAC (Trolox equivalent antioxidant capacity).

$$\mathsf{DPPH\; free\; radical\; scavenging\; rate}\left(\%\right)=\left(\mathsf{A0}-\mathsf{A}\mathsf{1}\right)/\mathsf{A}\mathsf{0}\times \mathsf{100}\%$$ (1)

2.2.2. ABTS radical scavenging capacity (ABTS)

The reaction mixtures were prepared with 10 μL of appropriately diluted sample solution and 190 μL of ABTS working solution. After thorough mixing, the samples were incubated in the dark at 25 °C for 6 min. The absorbance was measured at 734 nm. For the blank controls, the sample solution was replaced with extraction solution under identical conditions(Re et al., 1999). Assays were performed in sextuplicate. The ABTS radical scavenging activity was calculated according to Eq. (2) and expressed as TEAC:

$$\mathsf{ABTS}+\mathsf{inhibition\; rate}\left(\%\right)=\left(\mathsf{A0}-\mathsf{A}\mathsf{1}\right)/\mathsf{A}\mathsf{0}\times \mathsf{100}\%$$ (2)

2.2.3. Ferric-reducing antioxidant power (FRAP) assay

The FRAP working solution was prepared by mixing 0.3 mol/L acetate buffer (pH 3.8), 10 mmol/L 2,4,6-tripyridyl-s-triazine (TPTZ) working solution, and 20 mmol/L FeCl₃ solution at a volume ratio of 10:1:1. For the assay, 10 μL of appropriately diluted sample solution was combined with 190 μL of FRAP working solution (preheated to 37 °C). The mixture was allowed to react at room temperature (25 °C) for 10 min, after which the absorbance was measured at 593 nm. A blank control was prepared identically by replacing the sample solution with the extract solution . Each sample group was assayed across six replicates. The results are expressed as Trolox equivalents (μg/g dry weight (DW))(Benzie & Strain, 1996).

2.3. Untargeted metabolomic analysis

2.3.1. Sample extraction

Using vacuum freeze-drying technology, the wild jujube samples were placed in a lyophilizer (Scientz-100F), and then the samples were ground (30 Hz, 1.5 min) to powder form using a grinder (MM400, Retsch). Next, 50 mg of sample powder was weighed using an electronic balance (MS105DΜ), and 1200 μL of −20 °C-precooled 70% methanolic aqueous internal standard extract (less than 50 mg was added at a rate of 1200 μL of extractant per 50 mg of sample) was added. The mixture was vortexed once every 30 min for 30 s for a total of 6 times. After centrifugation (12,000 rpm, 3 min), the supernatant was aspirated, and the sample was filtered through a microporous membrane (0.22 μm pore size) and stored in an injection vial for UPLC–MS/MS analysis.

2.3.2. UPLC and MS conditions

Separation and identification of jujube metabolites were performed on an integrated liquid chromatography–mass spectrometry system. Chromatographic separation was carried out using a Waters ACQUITY UPLC system equipped with an HSS T3 column (1.8 μm, 2.1 × 100 mm) maintained at 40 °C. The mobile phase consisted of (A) 0.1% formic acid in water and (B) 0.1% formic acid in acetonitrile and was delivered at a flow rate of 0.40 mL/min. The gradient elution program was set as follows: 0–5.0 min, 95% A → 35% A; 5.0–6.0 min, 35% A → 1% A; 1% A was held until 7.5 min; subsequently, the composition was returned to 95% A within 0.1 min and equilibrated until 10.0 min. The injection volume was 4 μL, and the samples were injected in random order. Mass spectrometric analysis was conducted on a Sciex TripleTOF 6600 system equipped with an electrospray ionization source. The ion source parameters were set as follows: curtain gas, 35 psi; nebulizer gas, 50 psi; auxiliary heating gas, 60 psi; ion source temperature, 550 °C; declustering potential, ±80 V; and ion spray voltage, ±5500/−4500 V (positive/negative mode). Data were acquired in information-dependent acquisition mode. Full-scan MS spectra were collected over m/z 50–1250 with an accumulation time of 200 ms. Precursor ions with a charge state = 1+ and intensity exceeding 100 cps were selected for fragmentation, with up to 12 ions monitored per cycle and isotopic peaks within 4 Da excluded. The product ion scans covered the same m/z range of 50–1250 with an accumulation time of 40 ms, using a collision energy of 30 ± 15 eV (positive/negative mode) and unit resolution.

2.3.3. Metabolite quantification and data analysis

The raw mass spectrometry data were converted to mzXML format using ProteoWizard. XCMS was then employed for feature extraction, alignment, and retention time correction. Features with a missing data rate > 50% across experimental groups were filtered out. The remaining missing values were imputed using the K-nearest neighbors algorithm, and peak areas were normalized on the basis of quality control (QC) samples to correct for instrumental response drift. Metabolite identification was performed by matching against the MetWare in-house database, public databases, and predicted spectral libraries(Shen et al., 2019). Only high-confidence entries with a composite score > 0.5 and a QC coefficient of variation (CV) <0.3 were retained. For metabolites detected in both positive and negative ion modes, the entry with the highest confidence and lowest CV was kept after merging. To identify differentially abundant metabolites, the data were normalized and subjected to orthogonal partial least squares-discriminant analysis (OPLS-DA). Metabolites with a variable importance in projection (VIP) score > 1 and an absolute fold change ≥2 were defined as significantly differentially abundant metabolites.

2.4. Network pharmacology analysis and antioxidant target screening

All the metabolites identified in the untargeted metabolomics analysis were screened using the Traditional Chinese Medicine Systems Pharmacology Database and Analysis Platform (TCMSP; https://old.tcmsp-e.com/tcmsp.php). The screening criteria were set with reference to studies by Tao et al. (2013) and Cao et al. (2023) as oral bioavailability (OB) ≥ 20% and drug-likeness (DL) ≥ 0.10. Metabolites meeting these criteria were selected as potential active substances. Potential targets of the active compounds derived from wild jujubes were identified using the SwissTargetPrediction database (http://swisstargetprediction.ch/). Antioxidant-related targets were retrieved from the GeneCards database (https://www.genecards.org/). Common targets between the active compounds in jujubes and the antioxidant-related targets were subsequently identified. This intersection analysis was performed using the Venn diagram tool on the Bioinformatics Platform (http://www.ehbio.com/test/Venn/). KEGG pathway enrichment analysis was performed on the common targets using the DAVID database (https://david.ncifcrf.gov/). The ‘metabolite–target’ activity network and the core protein–protein interaction (PPI) network were constructed using Cytoscape software (version 3.9.1). Subsequently, topological analysis was performed on these networks. The core antioxidant targets were then identified on the basis of the following criteria: all five topological parameters—degree centrality (DC), betweenness centrality (BC), closeness centrality (CC), neighborhood connectivity (NC), and local average connectivity (LAC)—exceeded their respective average threshold values.

2.5. Molecular docking

Following network construction, core antioxidant targets demonstrating high relevance were identified through a two-stage screening process. This selection required all five topological parameters—DC, BC, CC, NC and LAC—to exceed their respective mean threshold values at both stages. The core antioxidant components were screened on the basis of a significant correlation with at least two antioxidant indicators or a DC greater than 10 in the ‘metabolite–target’ activity network. Molecular docking simulations were performed between the highly relevant core antioxidant targets and core antioxidant components. The three-dimensional protein structures of the core antioxidant targets were retrieved from the Protein Data Bank (PDB). The corresponding ligand structures of the active compounds were acquired from the PubChem database. All ligand structures were subsequently converted to PDB files using Open Babel software (version 2.3.1). The protein structures of the core antioxidant targets were subjected to hydrogen addition and water removal using PyMOL software (version 2.5.4). Molecular docking was then performed with AutoDock Tools (version 1.5.7) employing semiflexible docking protocols. The docking grid dimensions were optimized to encompass active sites prior to simulation. Finally, three-dimensional visualization of each of the docking poses was executed in PyMOL 2.5.4 (Zeng et al., 2025).

2.6. Statistical analysis

All the experiments were conducted with six biological replicates, and the data are presented as the mean ± standard deviation. The normality of the data distribution within each group was first assessed using the Shapiro–Wilk test. For data satisfying normality, one-way analysis of variance (ANOVA) followed by Tukey's post hoc test was applied; otherwise,thenon-parametric Kruskal–Wallis test with Dunn's post hoc test was used. All the statistical analyses were performed in SPSS 26.0. To further evaluate the power of the statistical tests, a post hoc power analysis was conducted on the basis of the observed effect size (e.g., partial η² or Cohen's d). This analysis was carried out using GPower 3.1 software, with the significance level (α) set at 0.05 and the statistical power (1-β) set at 0.8. Data visualization was performed using the MetWare Cloud Platform, Cytoscape 3.9.1, and Origin 2024.

3. Results and analysis

3.1. Untargeted Metabolomic data processing and annotation

3.1.1. Metabolomic analysis of wild jujube accessions

To elucidate metabolic profiles and their variations across wild jujubes from different provinces, we conducted untargeted metabolomic analysis using ultrahigh-performance liquid chromatography–triple quadrupole mass spectrometry (LC–QTOF–MS). The overlaid total ion current (TIC) chromatograms from the quality control (QC) replicates are shown in Fig. S1. High intersample TIC profile congruence (R²> 0.98) confirms the exceptional analytical reproducibility and data reliability. The detection of 2895 metabolites in dual ESI ± mode demonstrated the comprehensive metabolome coverage of the LC–QTOF–MS platform in wild jujube. This platform enables the comprehensive quantification of metabolites in six distinct wild jujubes grown in different provinces. Following peak extraction and alignment, 2895 metabolites were systematically identified across wild jujube fruit samples from diverse provinces using both positive and negative ionization modes. As illustrated in Fig. 2A, the metabolic profile of wild jujube comprises seven principal classes: amino acids and derivatives, organic acids, benzene derivatives, flavonoids, alkaloids, coumarins, phenolic acids, and terpenoids. Among these, amino acids and their derivatives (23.01%), organic acids (16.79%), and benzene and its derivatives (12.33%) were the main components of the metabolites, followed by flavonoids (5.35%), alkaloids (4.04%), coumarins (3.45%) and terpenoids (2.76%). These findings demonstrate the diversity and complexity of the metabolite composition in wild jujube fruits.

Fig. 2.

(A) Composition and abundance of metabolite classes identified in wild jujube fruit samples. (B) Hierarchical cluster analysis (HCA) dendrogram of six wild jujube accessions. (C) Principal component analysis (PCA) score plot of six wild jujube accessions. (D) Orthogonal projections to latent structures-discriminant analysis (OPLS-DA) score plot of six wild jujube accessions. (E) Permutation test validation (200 iterations) of the OPLS-DA model. (F) UpSet plot illustrating the distribution of differentially accumulated metabolites (DAMs) among pairwise comparisons of wild jujube from six provenances.

We performed unsupervised principal component analysis (PCA) on the samples (Fig. 2B). A clear separation trend was observed between different samples, indicating significant metabolic differences among the jujube accessions. Furthermore, replicate samples from each accession clustered tightly, demonstrating excellent reproducibility. The first two principal components (PCs) collectively accounted for 76.56% of the total variance, with PC1 contributing 67.8% and PC2 contributing 8.72%. Hierarchical cluster analysis (HCA) was applied to reduce metabolite dimensionality and extract categorical features (Fig. 2C). The 36 jujube samples were grouped into three distinct clusters. Samples from Shenyang formed a separate cluster. Samples from Gansu, Inner Mongolia, and Shaanxi exhibited similar profiles and clustered together. Similarly, samples from Hebei and Shanxi clustered together. Notably, the clusters containing samples from Gansu, Inner Mongolia, and Shaanxi were similar to the cluster containing samples from Hebei and Shanxi, which is consistent with the PCA results. To filter noise unrelated to classification and extract biologically relevant information from the metabolomic data (Nitta et al., 2017), we conducted orthogonal projections to latent structures discriminant analysis (OPLS-DA). As shown in Fig. 1D, compared with unsupervised PCA, OPLS-DA achieved greater separation between sample groups. To evaluate potential overfitting of the OPLS-DA model, we performed a 200-cycle permutation test (Fig. 2E). The permutation test yielded high R²Y (0.995) and Q²Y (0.988) values, confirming the excellent predictive ability, stability, and reproducibility of the model.

3.1.2. Metabolic variation in wild jujube accessions from different provenances

The OPLS-DA model was further employed to screen for differentially abundant metabolites in pairwise comparisons among wild jujube accessions from different provenances (SY, GS, TH, GL, NM, and HB). In all pairwise comparisons, both the Q² and R²Y values exceeded 0.9. These findings indicate that the models possessed good stability and reliability. The corresponding results are presented in Fig. S2. According to the OPLS-DA score plots (Fig. S2, A1-O1), distinct separation trends were observed between wild jujubes of different provenances. This clear separation indicates significant differences in metabolite composition among wild jujubes from different provenances. Differentially abundant metabolites were identified using the screening criteria of a VIP ≥ 1 and a fold change ≥2 or ≤ 0.5. These metabolites were visualized via volcano plots, as shown in Fig. S2 (A3-O3). A total of 15 pairwise comparison groups were analyzed: SY–GS, SY–TH, SY–GL, SY–NM, SY–HB, GS–TH, GS–GL, GS–NM, GS–HB, TH–GL, TH–NM, TH–HB, GL–NM, GL–HB, and NM–HB. The number of differentially abundant metabolites (DAMs) across the comparison groups is shown in Fig. 2F. Among all the comparisons, the TH vs. SY group exhibited the most significant metabolomic differences, with both the highest total number of DAMs (1503) and the greatest number of unique DAMs (69). These findings indicate that wild jujubes from the TH and SY provenances have the greatest differences in metabolite composition, potentially reflecting the most significant genetic or adaptive divergence. Furthermore, several comparison groups involving TH generally contained a greater number of DAMs, suggesting that the metabolome of wild jujube from the TH provenance may possess a greater degree of uniqueness. All the DAMs were merged and categorized, and the results are shown in Supplementary Fig. S3. Compared with the overall metabolite composition, the composition of the DAMs differed relatively little.

These results demonstrate substantial differences in metabolite composition and regulation among wild jujubes from different provenances. Fruit samples were collected under consistent environmental conditions, including site characteristics, ecological parameters, climatic factors, and cultivation practices. This standardized sampling method minimizes the influence of environmental variation on metabolite profiles. Consequently, the observed metabolic divergence is primarily attributable to inherent genetic characteristics. Heritable genetic variation has arisen during adaptation to environmental factors, including altitude, climate, soil type, and vegetation composition, across wild jujubes from different provenances. A study on dandelions (Bont et al., 2020) similarly demonstrated that long-term adaptation to distinct climatic conditions induced genetic variation in this species. Notably, significant differences in metabolite composition persisted even when all the populations were cultivated under identical environmental conditions. Similarly, research on Lycium ruthenicum (black goji berry) has revealed substantial genetic divergence among different provenances because of ecological variation. This genetic differentiation results in pronounced differences in amino acid composition among provenances (Li et al., 2023).

3.2. Elucidating antioxidant mechanisms in wild jujube fruit

3.2.1. Network pharmacology analysis of oxidative damage

Network pharmacology analysis represents an analytical approach that integrates systems biology and network analysis. This methodology involves the investigation of pharmacological processes and signaling networks centered on hub genes. It thereby elucidates the functions and behaviors of complex biological systems. Ultimately, network pharmacology enables the prediction of bioactive compounds targeting specific pathways and the exploration of their mechanisms of action(Zhai et al., 2025). Previous studies have established that wild jujube fruits possess potent antioxidant properties. They represent a valuable source of high-potency natural antioxidants (Wang et al., 2025). However, the metabolic composition of these fruits is significantly complex. Consequently, the specific bioactive compounds responsible for their antioxidant efficacy remain unclear. Furthermore, their mechanistic actions in mitigating oxidative damage require comprehensive investigation.

3.2.1.1. Prediction of active ingredients and core targets

TCMSP constitutes a specialized platform for herbal systems pharmacology. This database integrates comprehensive information on phytochemical constituents. It facilitates the analysis of disease–drug interaction networks through three core functions: (a) identification of bioactive compounds, (b) screening of drug targets, and (c) visualization of compound–target–disease networks. TCMSP serves as a pivotal resource for screening andcharacterizinghigh-potencyphytochemicals in medicinal plants (Liu et al., 2024).

We conducted comprehensive untargeted metabolomic profiling of wild jujubes from different provenances. After data integration and deduplication, 2895 distinct metabolites were identified. This constitutes the complete metabolic profile for wild jujube fruits. Subsequent screening via the TCMSP platform revealed 87 putative bioactive compounds. Furthermore, 1000 protein targets associated with these compounds were retrieved from the SwissTarget database. We identified 1302 antioxidant-related targets from the GeneCards database. Intersection analysis revealed 318 common targets shared between wild jujube fruits and antioxidant processes. To identify pivotal hub targets, we evaluated these 318 targets using the STRING database with a confidence score threshold >0.9. This screening yielded 273 potential antioxidant target proteins. We constructed a protein–protein interaction (PPI) network for these targets using Cytoscape software (Fig. 3B(a)). The resulting network comprised 273 nodes and 1230 edges, reflecting functional interactions and potential mechanisms. Core proteins were selected through topological analysis on the basis of five parameters: DC, BC, CC, NC, and LAC. All the values exceeded the average threshold. Two iterative screenings were performed. First, the core antioxidant target proteins were identified. Second, we aimed to identify highly correlated antioxidant core targets. The PPI network constructed from 41 core antioxidant targets is shown in Fig. 3B(b). This network comprises 41 nodes interconnected by 260 edges,41 core antioxidant targets and their topological characteristics are listed in Table 2. Seven hub antioxidant target proteins were identified: TP53, AKT1, SRC, STAT3, EP300, JUN, and ESR1. These targets are closely associated with oxidative damage and play crucial roles in a series of physiological and biochemical processes. For example, integrated network pharmacology and metabolomics analyses revealed that key antioxidant components in Dracaena cochinchinensis exert their antioxidant activity by modulating the AKT1 signaling pathway(Guo et al., 2024). Similarly, 28 core antioxidant compounds were identified in Siraitia grosvenorii (luohanguo) using a comparable approach; these compounds primarily function through key antioxidant targets such as GAPDH, AKT1, and TP53 (Liu et al., 2025). Numerous studies have demonstrated that TP53 is an important antioxidant target. It can reduce glycolysis and activate the pentose phosphate pathway, thereby providing protection against oxidative damage (AlMaazmi et al., 2025). AKT1 is another significant antioxidant target that is widely expressed across tissues. As a key downstream molecule in the PI3K signaling cascade, it regulates multiple pathways, including inhibition of apoptosis, stimulation of cell growth, and modulation of cellular metabolism (Liu et al., 2020). In summary, the antioxidant capacity of wild jujube fruit increases primarily by modulating the expression levels of seven key targets—TP53, AKT1, SRC, STAT3, EP300, JUN, and ESR1—along with other protein targets. These findings further support the value of wild jujube in the development of functional health foods and the extraction of bioactive compounds.

Fig. 3.

(A) UpSet plot showing intersections between antioxidant-related targets and bioactive compound targets across six wild jujube accessions. (B) Protein–protein interaction (PPI) networks: (a) network of 273 high-confidence common targets; (b) core subnetwork of 41 screened hub targets. (C) Compound–target interaction network. (D) Gene Ontology (GO) enrichment analysis of 273 high-confidence common targets. (E) KEGG pathway enrichment analysis of 273 high-confidence common targets.

Table 2.

Topological characteristics of 41 core antioxidant targets.

No.	Target	Betweenness	Degree	Eigenvector	LAC	Network
1	TP53	10,688.8175	64	0.2630	9.0000	39.2805
2	HSP90AA1	4245.3290	41	0.1671	6.0976	17.5508
3	HSP90AB1	1563.7999	30	0.1275	5.5333	13.5215
4	CTNNB1	1843.6643	33	0.1706	7.8182	15.4448
5	ESR1	4097.1915	35	0.2136	11.3143	19.6746
6	AKT1	4411.8318	42	0.2029	8.2381	20.3060
7	MAPK1	1813.9865	33	0.1781	9.2121	16.5681
8	CCND1	940.5156	23	0.1238	7.3913	11.1576
9	MAPK14	751.0974	20	0.1019	6.0000	10.0654
10	MAPK3	1840.6111	32	0.1648	8.5000	15.4420
11	EP300	3925.7064	33	0.1545	8.0606	17.1443
12	PIK3CA	1989.4423	33	0.1417	7.0909	16.5248
13	STAT3	2938.3366	40	0.2154	9.6500	19.5063
14	CASP3	2415.5236	23	0.0956	5.2174	9.7794
15	JUN	4405.7628	35	0.1866	9.7143	18.2474
16	SRC	3678.4265	42	0.2102	9.5238	23.2172
17	BCL2	784.9579	22	0.1276	7.6364	10.7672
18	TNF	2838.0690	26	0.0951	5.8462	12.8903
19	HIF1A	2315.0333	20	0.1360	7.7000	9.2782
20	JAK2	750.5314	20	0.0844	7.8000	11.8938
21	JAK3	612.6643	13	0.0500	3.8462	6.0417
22	HRAS	1986.3948	29	0.1315	6.6207	12.2832
23	MAPT	834.8418	15	0.0580	3.2000	4.9190
24	PRKACA	1615.0162	28	0.1245	5.0000	8.6638
25	MAPK9	672.5945	25	0.1203	6.1600	11.9447
26	MAPK8	2334.2017	29	0.1400	6.5517	13.3341
27	PRKCG	693.6597	13	0.0533	5.6923	6.9000
28	PRKCA	2199.8041	19	0.0670	4.8421	8.9383
29	PRKCB	1406.4732	16	0.0576	5.2500	6.8828
30	RELA	856.1399	25	0.1398	8.1600	11.2473
31	TLR4	1246.1258	16	0.0569	3.6250	4.7444
32	HSPA8	905.1269	17	0.0760	4.7059	8.7649
33	CXCL2	664.4810	11	0.0366	5.2727	7.8556
34	CXCL8	1170.7395	15	0.0430	5.6000	9.6302
35	SIRT1	699.2943	15	0.0748	4.6667	6.6071
36	KDR	998.7672	15	0.0574	4.5333	7.4745
37	IL6	2626.2261	25	0.0987	5.9200	11.5210
38	NFKB1	1282.6791	23	0.1260	7.6522	10.3234
39	EGFR	1816.7690	28	0.1463	9.0714	15.4639
40	ESR2	583.7876	11	0.0733	6.0000	6.7917
41	PTK2	1510.0981	23	0.1055	6.1739	10.1020

To investigate interactions between bioactive compounds and core antioxidant targets, we constructed a component–target protein–protein interaction (PPI) network. This network integrated 41 core antioxidant target proteins with 77 associated bioactive compounds. The resulting network comprised 118 nodes (77 compound nodes and 41 protein nodes) connected by 503 edges(Fig. 3c). A screening based on a DC >10 identified 12 key bioactive compounds. These findings align with the fundamental polypharmacology synergy principle of network pharmacology, in which multiple components interact with multiple targets. Phytochemicals in dietary contexts primarily exert bioactivity through synergistic effects among multiple compounds. For instance, epigallocatechin gallate (EGCG) and kaempferol cooperatively upregulate SOD, CAT, and GSH-Px activities. This enzymatic activity increases the efficiency of ROS scavenging, thereby balancing cellular oxidative stress (Zhang et al., 2023). Similarly, the synergistic effects of resveratrol and curcumin on hypertension management have been demonstrated. Their combined antioxidant and antiangiogenic effects effectively reduce blood pressure (Hesarooeyeh et al., 2024).

3.2.1.2. Functional enrichment analysis of GO and KEGG pathways

To elucidate the antioxidant mechanisms of wild jujube fruit, we conducted Gene Ontology (GO) enrichment analysis on 41 core antioxidant targets using DAVID. The top 10 significantly enriched terms are presented in Fig. 3D. Biological processes (BPs) were predominantly associated with negative regulation of gene expression, positive regulation of RNA polymerase II transcription, signal transduction, negative regulation of the apoptotic process, and related pathways. Molecular functions (MFs) were significantly enriched in enzyme binding, protein kinase activity, ATP binding, protein kinase binding, and identical protein binding. The cellular component (CC) terms were enriched primarily in the following: cytosol, nucleoplasm, nucleus, cytoplasm, and mitochondria. These findings demonstrate that compounds in wild jujube alleviate oxidative damage through the coordinated modulation of BPs across multiple cellular compartments and functions. KEGG enrichment analysis revealed 179 oxidative stress-associated pathways (Fig. 3E). Among the top 20 significantly enriched pathways, the key entries included pathways associated with cancer, lipids and atherosclerosis, hepatitis B, Kaposi sarcoma-associated herpesvirus infection, and proteoglycans in cancer. Notably, wild jujube metabolite targets were significantly enriched in cancer-related pathways. These findings suggest that its antioxidant components may reduce cancer incidence or suppress tumor progression. Research has shown that oxidative stress critically contributes to carcinogenesis and cancer progression (Luo et al., 2022). Elevated ROS levels increase cancer risk when compromised antioxidant systems fail to protect cells from oxidative damage (Luo et al., 2022). These findings validate the significant enrichment of wild jujube metabolite targets associated with cancer-related pathways. Concurrently, significant enrichment was observed for both the proteasome pathway and the phosphatidylinositol 3-kinase/protein kinase B (PI3K/AKT) signaling pathway. As a key regulator of cellular oxidative stress and survival, PI3K/AKT signaling critically enhances cellular antioxidant defenses and neuroprotection. This pathway is associated with cancer, inflammation, and related pathologies (Fruman et al., 2017; Goyal et al.,2023). Therefore, wild jujube fruit holds considerable promise for preventing and treating these diseases.

3.2.1.3. Evaluation of antioxidant capacity and screening of key antioxidant compounds in wild jujube from different provenances

Chemical antioxidant assays represent among the most prevalent and accessible methods for evaluating sample antioxidant capacity (Munteanu & Apetrei, 2021). Wild jujube, a notable medicinal and edible plant, exhibits significant antioxidant properties that effectively mitigate oxidative damage. In this study, antioxidant capacity was quantified through multiple established assays: DPPH radical scavenging, ABTS inhibition, and ferric reducing antioxidant power (FRAP). These complementary approaches enabled a comprehensive cross-assessment of variations in antioxidant capacity across geographical provinces. The results of the chemical antioxidant assays are presented in Fig. 4A. The DPPH radical scavenging capacity ranged from 242.95 ± 10.03 to 289.55 ± 11.21 μg TE/g DW.The ABTS inhibition values varied between 1655.09 ± 71.95 and 1965.61 ± 60.87 μg TE/g DW. The ferric reducing antioxidant power (FRAP) ranged from 17.24 ± 0.72 to 20.25 ± 0.41 μg TE/g DW. Significant differences (p < 0.05) in antioxidant capacity were observed across wild jujubes from different provinces. However, fruits from the NM and HB provinces differed significantly only in terms of their FRAP values but presented comparable DPPH and ABTS activities. Post hoc poweranalysis confirmed that the current sample size(n = 6 per province) achieved a statistical power ≥ 0.80 for detectingthe observed effect sizes in all three antioxidant assays; detailed results are provided in Supplementary Table S3.

Fig. 4.

(A) Free radical scavenging capacities across wild jujube accessions: (a) ABTS, (b) DPPH, and (c) FRAP assays. (B) Correlations between antioxidant indicators and 41 bioactive compounds. The colored lines denote positive correlations; the gray lines indicate negative correlations.

The Pearson correlation coefficient method has been successfully applied to screen for antioxidant components in red wine (Peng et al., 2024). In the current study, we calculated the Pearson correlation coefficients between antioxidant indices and 41 compounds associated with seven key antioxidant targets in wild jujube. These correlations were visualized as a network heatmap(Fig. 4B). This approach facilitates the comprehensive identification of the factors influencing the antioxidant activity of wild jujube. The results were visualized as a network heatmap, enabling comprehensive identification of factors influencing antioxidant activity. The results demonstrated that eupatilin, licoisoflavone A, liquiritin, deserpidine, Wuweizisu C, and paucin were highly significantly positively correlated (p < 0.01) with the results of the ABTS, DPPH, and FRAP assays. Liquiritin demonstrated the strongest correlation. Liquiritin, a flavonoid compound, is one of the primary active constituents of licorice. Treatment with liquiritin has been shown to increase the expression of peroxiredoxin-6 (Prdx6) in mice, thereby alleviating oxidative damage (Liu et al., 2022). Additionally, liquiritin reduces mercury uptake in Arabidopsis under Hg stress while increasing peroxidase (POD) activity and glutathione (GSH) synthesis. These effects mitigate oxidative injury and support normal growth in Arabidopsis (Gao et al., 2024). Thus, liquiritin is a high-quality antioxidant. In this study, a total of 18 core antioxidant components were identified on the basis of the criterion of a significant correlation with at least two antioxidant indicators. Among these, the antioxidant properties and mechanisms of several compounds—such as 6-gingerol, quercetin, procyanidin B1, ganoderic acid A, and thalictrine—have been previously confirmed. However, the antioxidant activities of other compounds, including Paucin, cymarin, and asparagoside A, remain less studied.

3.2.1.4. Comparative analysis of core antioxidant component levels across six provenances

To elucidate the material basis underlying the differences in the antioxidant capacity of wild jujube from different geographical origins, this study performed a quantitative comparison of the relative abundance of the 26 screened core antioxidant components across six provenances on the basis of chromatographic peak areas (Supplementary Fig. S4). Statistical analysis revealed significant differences (P < 0.05) in the relative contents of these core components among the provenances, demonstrating a clear provenance-specific accumulation pattern. Several components exhibited marked geographical specificity. For instance, medicarpin had the greatest peak area in samples from the GL provenance, which was nearly twice that observed in samples from other provenances. Conversely, although eupatilin was present in most provenances, its peak area was nearly undetectable in TH samples. Liquiritin displayed the opposite accumulation trend, with a significantly higher content in the TH samples than in the others, while it was almost undetectable in the HB samples. Furthermore, the peak areas of components such as cymarin, wuweizisu C, and ganoderic acid A significantly differed among provenances, which may explain the significant variation in antioxidant capacity across samples. In contrast, bergaptol was commonly present in all samples, with minor quantitative differences; quercetin was also widely present, showing significant differences only between the SY provenance and the others. In summary, the 26 core antioxidant components displayed distinct accumulation patterns in wild jujube from different geographical origins. Key components such as medicarpin, liquiritin, ganoderic acid A, and isomucronulatol exhibited obvious geographic clustering, whereas components such as bergaptol and quercetin were universally present across samples and thus could not serve as characteristic markers for distinguishing provenances. These findings not only provide direct metabolomic evidence explaining the differences in antioxidant activity among wild jujube from different origins but also clarify the crucial regulatory role of geographical origin in the synthesis of secondary metabolites in wild jujube.

3.3. Molecular docking of core antioxidant components and core antioxidant targets

Molecular docking serves as a critical methodology for elucidating interactions between bioactive compounds and receptor proteins (Shoily et al., 2025). This approach is extensively employed in drug discovery and bioactive compound identification(Chikowe et al., 2024). To further investigate the antioxidant mechanisms of wild jujube, we performed molecular docking using AutoDock Vina. Twenty-six core antioxidant components were docked against seven hub antioxidant targets: TP53, AKT1, SRC, EP300, STAT3, JUN, and ESR1. The binding site parameters for validating the docking reliability of ligand–target complexes are shown in Supplementary Table S4. The results are presented in Fig. 5A. Lower binding energy values indicate stronger molecular interactions. Binding affinities ≤ −5 kcal/mol demonstrate effective ligand–protein binding. All 26 wild jujube compounds exhibited binding energies < −5 kcal/mol against six hub targets. This result confirms the strong binding between core antioxidants and their molecular targets. Notably, the cymarin–ESR1 (−11.3 kcal/mol), procyanidin B1–SRC (−10.8 kcal/mol), and licoisoflavone A–JUN (−9.9 kcal/mol) complexes demonstrated exceptional binding affinities.

Fig. 5.

(A) Heatmap of binding affinities (kcal/mol) between bioactive compounds and antioxidant targets in molecular docking studies. (B—D) Representative binding pose visualizations of (B) the cymarin–ESR1 complex, (C) the procyanidin B1–SRC complex, and (D) the licoisoflavone A–JUN complex.

Cymarin forms hydrogen bonds with ARG-99 and THR-15 of ESR1, whereas procyanidin B1 forms hydrogen bonds with ARG-99 and THR-15 of SRC. Licoisoflavone A forms hydrogen bonds with LYS-298, ASN-394, and ALA-393 of JUN; these specific interactions increase binding stability. The 3D representations of the molecular docking results are shown in Fig. 5 (B, C, D). As a cardiac glycoside, cymarin primarily functions by targeting and regulating the expression of the PAX6 gene, making it a potential candidate drug in the field of breast cancer treatment. Furthermore, the PAX6 gene plays a critical role in cellular resistance to oxidative stress (Li et al., 2023; Zheng et al., 2022). However, cymarin belongs to the cardiac glycoside class, whose therapeutic window is narrow, and excessive intake can easily lead to adverse reactions such as cardiotoxicity. Its application must be strictly controlled under medical supervision. Procyanidin B1 is a natural dimeric flavonoid with significant antioxidant activity. When grape-processing byproducts are used as the main source of polyphenolic compounds in mixed rye bread, both the procyanidin B1 content and the antioxidant activity of the bread increase (Mildner-Szkudlarz et al., 2011). Current research indicates that this compound is highly safe, with no significant side effects observed at conventional intake levels, making it a valuable natural antioxidant for development(Zeng et al., 2020). As an isoflavone, in vitro studies have confirmed that licoisoflavone A can effectively scavenge free radicals, alleviate oxidative stress, and protect cells from oxidative damage (Quesada et al., 2012). Additionally, research has shown that licoisoflavone A enhances the deacetylation level of superoxide dismutase 2 (SOD2) by activating the deacetylase Sirt3, thereby increasing its antioxidant activity (Guo et al., 2020). Currently, toxicological studies on this compound are insufficient. However, on the basis of the common characteristics of isoflavones, the potential effects on the endocrine system under high-dose or long-term intake conditions require attention (Wang et al., 2007). In summary, the aforementioned compounds all demonstrate promising antioxidant potential, but their practical application prospects require comprehensive evaluation on the basis of their respective activity profiles and safetyconsiderations.

ESR1 can competitively bind to the DGR domain of the Keap1 protein via its highly conserved C-terminal DLL motif. This interaction blocks Keap1, which acts as anadaptor for the E3 ubiquitin ligase complex, from promoting the ubiquitination and degradation of the transcription factor Nrf2. Consequently, the Nrf2 protein is rapidly stabilized, leading to activation of the Nrf2–ARE pathway. This results in the systemic upregulation of endogenous antioxidant enzymes such as HO-1 and SOD, thereby increasing the ability to scavengeROS(Yang et al.,2023). This mechanism has been validated in both estrogen-deficient osteoporosis and glucocorticoid-induced osteonecrosis of the femoral head models(Zhu et al., 2025). Thus, ESR1 plays a central role in maintaining redox homeostasis in bone tissue. Moreover, membrane-associated ERα can rapidly activate the PI3K/Akt/eNOS and MAPK/ERK pathways, promoting NO production within minutes and transmitting immediate survival signals. Additionally, ERα is enriched at mitochondrial-associated membranes (MAMs), where it stabilizes the electron transport chain, reduces electron leakage, and upregulates mitochondrial antioxidant enzymes such as Mn-SOD through the Nrf2/PGC-1α axis, thereby mitigating oxidative damage at its source(Simoncini et al., 2006). The core function of the JUN protein in regulating oxidative stress lies in suppressing the overactivation of the JNK/JUN/AP-1 signaling pathway, which serves as a key amplifier of oxidative stress in inflammatory, stress, and immune responses. Upon cellular stimulation, JNK is phosphorylated and activates JUN, which then forms the AP-1 transcription complex with proteins such as c-Fos. After entering the nucleus, AP-1 initiates the expression of multiple pro-oxidant and proinflammatory genes, including iNOS, COX-2, TNF-α, and IL-6, thereby exacerbating oxidative damage (Kumar et al., 2024). Therefore, inhibiting the activation of JNK/JUN effectively blocks the AP-1-mediated transcription of pro-oxidant genes, reducing the generation of oxidative stressors at the source. Concurrently, such inhibition alleviates the suppression of endogenous antioxidant systems, promoting the expression and restoring the activity of antioxidant enzymes such as superoxide dismutase (SOD), catalase (CAT), and glutathione peroxidase (GPx), thereby enhancing the overall antioxidant defense capacity of the cell(Kim et al., 2000). This mechanism has been notably demonstrated in response to various phytochemicals, such as physalin A and polyphenols. Hence, the JUN protein plays a critical role in regulating the homeostasis of the oxidative system(Saleem et al., 2025). SRC, a nonreceptor tyrosine kinase, serves as a key regulatory node in the redox signaling network within the cardiovascular system. It is not itself a direct ROS-scavenging antioxidant protein but participates in the initiation and amplification of oxidative stress through a bidirectional interplay mechanism. On the one hand, SRC can be activated by stimuli such as angiotensin II, subsequently phosphorylating p47phox or indirectly enhancing the activity of NADPH oxidases (particularly Nox1, Nox2, and Nox5) via kinases such as PKC, thereby promoting ROS production(Camargo et al., 2022; Touyz et al., 2003). On the other hand, ROS (especially H₂O₂) can oxidatively modify specific cysteine residues on SRC (Cys277 and Cys185), inducing conformational changes, promoting autophosphorylation, and weakening inhibitory interactions, thereby creating a positive feedback loop that activates SRC(Heppner et al., 2018). This bidirectional ‘SRC–ROS’ signaling pathway continuously regulates the activation of downstream pathways such as MAPK and PI3K/Akt, collectively contributing to the activation of antioxidant systems. Therefore, SRC can be considered a central signaling hub that links oxidative stress to pathological damage and plays a crucial role in regulating ROS homeostasis (Hussain et al., 2023).

In summary, cymarin, procyanidin B1, and Licoisoflavone A serve as core antioxidants in wild jujube, whereas ESR1, SRC, and JUN play essential roles as core target proteins within the antioxidant system. However, the current research is limited to a theoretical framework based on data mining and molecular docking analyses. Further cellular assays are essential to validate the antioxidant mechanisms of these natural compounds and assess their potential applications in the food industry.

4. Conclusion

In this study, LC–QTOF–MS-based untargeted metabolomics was used to analyze differences in metabolite composition among wild jujube fruits from six different provinces, and the core antioxidant components and their mechanisms of action were further investigated using network pharmacology and molecular docking. The results revealed significant differences in metabolic profiles among fruits from different provinces. Network pharmacology analysis revealed that 26 key antioxidant compounds in wild jujube primarily exert antioxidant effects by regulating seven key antioxidant targets through pathways related to cancer, lipid metabolism, and atherosclerosis. These 26 key antioxidant compounds can form stable hydrogen bonds with the seven core targets, among which the cymarin–ESR1, procyanidin B1–SRC, and isolicoflavonol A–JUN complexes exhibited the lowest binding energies. This study clarifies the metabolic differences among jujube samples from different provenances and provides an in-depth exploration of their active antioxidant substances and mechanisms of action. These findings offer fundamental data to support the extraction of active ingredients and the development of functional health products from jujube as a medicinal and edible plant and lay a foundation for elucidating the physiological basis and clinical applications of its bioactive components.

CRediT authorship contribution statement

Pengfei Chen: Writing – original draft, Methodology, Investigation, Formal analysis, Data curation. Shengjun Dong: Visualization, Investigation, Formal analysis. Ling Chu: Software, Formal analysis, Data curation. Lei Zhang: Visualization, Investigation, Data curation. Wei Qin: Writing – review & editing, Validation, Investigation, Funding acquisition.

Funding

This study was financially supported by Autonomous Region Major Science and Technology Special Project(2023A02010–01), Autonomous Region Rural Revitalization Talent Program (2023SNGGGGCC012),and Xinjiang Hetian College Industry–University Collaborative “Merit-Based Talent Recruitment” Project(2025JSFW-02).

Declaration of competing interest

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

Data availability

Data will be made available on request.