Methanol and formic acid (FA) were purchased from Sigma, and MS–grade acetonitrile was purchased from Merck. Fresh L. ruthenicum fruits were collected in Qinghai and Xinjiang Provinces, China.

The fruits were lyophilized and subsequently ground into a fine powder. For extraction, 200 mg of the powder was subjected to ultrasound-assisted extraction with 10 mL of 75% methanol for 30 min. After extraction, the mixture was centrifuged at 5000× g for 10 min, and the supernatant was filtered through a 0.22 μm filter prior to UPLC-QTOF/MS analysis. Three replicates were prepared for each sample to assess consistency.

UPLC-QTOF/MS analysis was performed using a Waters system coupled with a QTOF mass spectrometer (Waters Corporation, Milford, MA, USA). The flow rate was set at 0.3 mL/min. The chromatographic separation was achieved using a UHPLC column (250 mm × 2.1 mm, 1.8 μm). The mobile phase consisted of phase A (water with 0.1% formic acid, FA) and phase B (acetonitrile with 0.1% FA), with the following gradient elution: 3%–15% B for 0–20 min, 15%–30% B for 20–40 min, 30%–95% B for 40–45 min, and 95% B for 45–55 min. Peak identification was performed using a PDA detector set to 280 nm, with UV–vis spectra acquired from 190 to 600 nm.

The mass spectrometry parameters were as follows: Electrospray ionization (ESI) in negative ion mode, source temperature of 100°C, cone gas flow of 50 L/h, desolvation temperature of 400°C, capillary voltage of 2.50 kV, mass range of 50–1500 m/z, and a scan rate of 0.25 s. High-resolution mass data were collected at both high and low collision energies, and both fragments and precursor ions were analyzed simultaneously.

The raw data obtained from the samples were processed using Progenesis QI software (Nonlinear Dynamics, UK) for peak detection and automatic alignment. Compound identification was initially performed by querying the METLIN database, followed by further confirmation through the comparison of MS/MS fragmentation patterns and accurate masses. Where possible, identifications were cross-referenced with known standards and existing literature. To further validate the results, reference compounds were employed, and the findings were compared with previously reported data on similar compounds. Multivariate statistical analysis was conducted using SIMCA-P 14.1 (Umetrics, Sweden), which included Principal Component Analysis (PCA) and Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA).

The secondary metabolites identified through UPLC-QTOF/MS analysis were imported into the Swiss Target Prediction database (http://SwissTargetPrediction) to predict the targets of each compound. The resulting targets were then merged for further analysis. To identify target proteins associated with inflammation, the keyword “inflammation” was searched in the DRUGBANK and DisGeNET databases, and gene names were retrieved from the UniProt database. The candidate component targets and disease-related targets were imported into the STRING platform to construct a protein–protein interaction (PPI) network. This PPI network was then analyzed using Cytoscape and the NetworkAnalyzer tool for network topology. The degree values of the nodes were calculated to rank the genes, with genes having degree values above the average threshold considered key targets. In total, 64 key targets were identified and further analyzed for their potential roles within the network. These key targets were imported into the DAVID database for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses to obtain biological information and explore potential pharmacological mechanisms of action.

Under the UPLC conditions described above, 12 samples from wild and cultivated plants were analyzed. Peaks consistently observed across all chromatograms were assigned, demonstrating the reproducibility and reliability of the UPLC separation. Fig. 1a shows the UPLC chromatographic profile at 280 nm, highlighting the separation of metabolites based on retention time. In contrast, Fig. 1b presents the total ion chromatogram (TIC) acquired in negative ion mode, offering an overview of the detected metabolites and their relative intensities in mass spectrometry.

The validation of the UPLC-QTOF/MS analytical method enabled the identification and authentication of compounds in L. ruthenicum. A total ion chromatogram in negative ESI mode under optimal conditions is shown in Fig. 1b, with compound formulas, accurate masses, fragment ions, and retention times of the major peaks listed in Table 1. To ensure the reliability of metabolite identification, the mass accuracy (expressed as ppm error) and identification confidence levels were included. A total of 62 peaks were detected, primarily consisting of alkaloids, anthocyanins, phenolic acids and derivatives, and flavonoids.

Peaks 1, 6, 9, 11, 15, 17–19, 24, 26, 27, and 32 were identified as phenolic acids and derivatives. MS data showed ions at m/z 191 ([quinic acid–H]⁻) and 179 ([caffeic acid–H]⁻), suggesting the presence of quinic acid and caffeoyl moieties, respectively. Peak 1 was identified as quinic acid, peaks 9 and 24 as chlorogenic acid isomers, and peaks 15, 16, 17, and 25 as dicaffeoylquinic acid isomers [22]. Peak 4 was identified as protocatechuic acid glucoside, with a protonated molecular ion at m/z 315.0716 and fragment ions at m/z 153 (protocatechuic acid) and m/z 162 (glucoside) [23].

A total of 11 alkaloids (10, 12, 23, 30, 37, 39, 42, 43, 44, 47, and 48) were detected. Tryptophan (peak 10) showed a protonated molecular ion at m/z 203.2083, with fragment ions at m/z 159, 142, and 116 ([M–H–CO₂]⁻, [M–H–NH₃]⁻, [M–H–CO₂–NH₃–C₂H₂]⁻) [21]. Peak 45, with a parent ion at m/z 245.0928, was identified as N–acetyl–tryptophan [24]. Peak 23 matched the spectrum of N,N–bis(dihydrocaffeoyl) spermine from L. ruthenicum [4]. Peaks 28 and 31, with a parent ion at m/z 794.3343, were identified as N–caffeoyl, N–dihydrocaffeoyl spermidine dihexose isomers [25]. Peak 32 was identified as lycibarbarspermidine H, based on the published spectrum from L. barbarum [26]. Peaks 40, 41, and 42 were identified as caffeoyl (dihydrocaffeoyl) spermidine isomers ([M–H]⁻ at m/z 470), while peak 35, with a parent ion at m/z 632, was identified as dihydrocaffeoyl-caffeoyl spermidine hexose, matching the fragmentation pattern of peak 42. Peak 37, with a parent ion at m/z 472.2452, was identified as N,N–bis-dihydrocaffeoyl spermidine [4]. Peak 48, with a parent ion at m/z 468.2131, was identified as N,N–dicaffeoylspermidine [4].

Fourteen flavonoid compounds were identified (peaks 27, 34, 36, 39, 47, 48, 49, 51–54, 58, 60, 61). Peak 36, with a parent ion at m/z 771.1979 ([M–H]⁻), produced fragment ions at m/z 609 and 301, which indicated it was quercetin-rutinoside-hexose [27]. Peaks 39, 48, 58, and 61 were identified as isorhamnetin-glucoside-gentiobioside, kaempferol-rutinoside-glucoside, isorhamnetin 3-rutinoside, and kaempferol-caffeylsophoroside, respectively, based on coelution with standards and MS confirmation. Peak 47, with a parent ion at m/z 625.1405, was identified as myricetin 3-rutinoside [28]. Peak 49, with a parent ion at m/z 579.1717 and a fragment ion at m/z 271, was identified as naringin. Peak 51, with a parent ion at m/z 597.1815, was identified as phloretin-diglucoside [29]. Peak 52, with a parent ion at m/z 493.0983 and a typical fragment at m/z 331, was identified as laricitrin 3-glucoside. Peak 60, with a parent ion at m/z 433.1131, was identified as helichrysin A, matching a previous study on L. ruthenicum [3].

Anthocyanins, which contribute to the plant’s vibrant coloration, were also abundant in L. ruthenicum. Eight anthocyanins (peaks 29, 43, 44, 50, 55–57, and 59) were identified by matching their mass spectra with those reported in previous studies of L. ruthenicum [7]. These included several 3,5–diglycoside derivatives of petunidin, acylated with phenolic acids, as well as trans–cis isomeric forms.

To compare the metabolomic profiles among wild and cultivated L. ruthenicum fruits, multivariate analyses, including PCA and OPLS–DA, were performed on the MS data. Ion features (retention time, m/z) for L. ruthenicum were extracted from the raw MS data. The PCA score plots are shown in Fig. 2, with clear separation of three distinct groups in the PC1/PC2 score plot. These groups corresponded to the wild Qinghai, wild Xinjiang, and cultivated Qinghai samples. This separation indicates that the chemical profile of L. ruthenicum is a significant factor influencing fruit quality.

OPLS–DA was employed to compare the chemical profiles of L. ruthenicum from different collection and cultivation sites. The R²Y value of the model was 0.983, and the Q² was 0.957, indicating that the model was well established and had strong predictive ability [8,30]. The OPLS–DA score chart (Fig. 3) clearly classified the samples into three distinct groups, consistent with the PCA results, based on their chemical profiles.

Potential sources of variation that could influence the results of the multivariate analysis include biological factors (such as differences in fruit maturity and environmental conditions) and inconsistencies in sample preparation. To minimize these variations, all samples were processed following the same protocol and analyzed under identical instrumental conditions. These measures effectively reduced the impact of experimental noise and enhanced the robustness of the model.

The identification of compounds with high VIP values provides valuable insights into the metabolites that drive the differentiation of L. ruthenicum samples. Compounds 30, 35, 39, 41, 53, and 55, based on their accurate mass data and retention times, serve as key markers for distinguishing between the different samples. These findings suggest that these specific compounds can be utilized not only to authenticate L. ruthenicum but also to potentially monitor quality and assess variations in plant samples from different origins or cultivation practices. The use of these compounds in further studies could help refine the authentication process and provide additional insights into the metabolic profile of L. ruthenicum.

The interplay between ecological factors and cultivation practices plays a pivotal role in shaping the chemical composition and overall quality of L. ruthenicum. The distinct environmental conditions found in regions such as Qinghai and Xinjiang provide a natural basis for variations in metabolite profiles. Qinghai’s plateau environment, with its prolonged sunshine durations and saline soils, offers an ideal setting for L. ruthenicum to flourish, which may contribute to the superior quality of the plants produced there [31,32]. Furthermore, the differences in longitude, latitude, altitude, and precipitation between Qinghai and Xinjiang further influence the chemical composition of L. ruthenicum, underscoring the complexity of plant responses to their environmental context. As noted, cultivation practices such as fertilizer use, irrigation, and farming techniques also play a critical role in shaping the metabolite profiles of the plant. These anthropogenic factors can lead to distinctions between wild and cultivated L. ruthenicum, a trend that is clearly reflected in the PCA and OPLS-DA models, where samples from these two groups were distinctly segregated [33].

These observations suggest that a holistic approach, integrating both ecological and cultivation factors, is essential for understanding and optimizing the quality of L. ruthenicum. Adjusting cultivation practices to replicate the ideal environmental conditions may enhance the medicinal properties and chemical composition of cultivated plants, offering a promising avenue for improving consistency and quality in production.

A total of 44 anti-inflammatory metabolites were identified from the 59 metabolites obtained, and 247 potential targets were selected based on a database search. Using intersecting targets, the STRING database was employed to construct a protein–protein interaction (PPI) network, as shown in Fig. 4. This network comprises 247 nodes and 691 edges, with an average degree of 5.6. The node size is proportional to its degree, and the intensity of the node color correlates with its degree, where a redder hue indicates a higher value. The edge thickness and color represent the combined scores between interacting nodes (Fig. 4). Topological analysis identified several key anti-inflammatory targets, including SRC (degree = 41), STAT3 (degree = 40), MAPK1 (degree = 37), and HSP90AA1 (degree = 37), all of which exhibit significant roles within the network, highlighting their importance in the anti-inflammatory effects of L. ruthenicum.

GO enrichment analysis revealed a total of 2497 biological processes, 163 cellular components, and 226 molecular functions. The top ten terms from each category are illustrated in Fig. 5. The most prominent biological process was the response to molecules of bacterial origin, followed by the response to lipopolysaccharide and regulation of the inflammatory response. In the cellular component category, the most significant terms included membrane raft, membrane microdomain, and membrane region. In molecular function, endopeptidase activity, DNA-binding transcription factor binding, amide binding, and protein serine/threonine kinase activity ranked highest.

KEGG enrichment analysis identified 156 signaling pathways, with the top twenty pathways listed in Fig. 6. Among the inflammation-related pathways, the top ten pathways exhibiting the highest degrees of enrichment included the lipid and atherosclerosis pathway, neuroactive ligand-receptor interaction pathway, calcium signaling pathway, microRNAs in cancer pathway, chemical carcinogenesis-receptor activation pathway, MAPK signaling pathway, human cytomegalovirus infection pathway, proteoglycans in cancer pathway, and PI3K-Akt signaling pathway.

As illustrated in Fig. 7, Cytoscape software was used to map the relationships between metabolites, targets, pathways, and diseases. In the map, red represents the targets of L. ruthenicum metabolites, blue denotes the metabolites themselves, green highlights the top twenty pathways, and yellow represents the associated diseases. The anti-inflammatory effects of L. ruthenicum are likely mediated through each active metabolite, its related targets, and the corresponding pathways. This study comprehensively captured the multimetabolite, multitarget, and multipathway mechanisms underlying the anti-inflammatory action of L. ruthenicum.

As one of the most prominent compounds in L. ruthenicum, anthocyanins are widely recognized for their anti-inflammatory properties. These metabolites modulate the MAPK and NF-κB signaling pathways, which are crucial in regulating pro-inflammatory cytokines such as TNF-α and IL-1β. Anthocyanins inhibit the activation of MAPKs, reducing the production of inflammatory cytokines and alleviating inflammation-related diseases. Additionally, phenolic acids, including caffeic acid and chlorogenic acid, were identified as significant contributors to the anti-inflammatory effects of L. ruthenicum. These compounds inhibit the expression of inflammatory mediators by suppressing the NF-κB signaling pathway, reducing the secretion of pro-inflammatory cytokines. Phenolic acids also possess antioxidant properties, further enhancing their anti-inflammatory activity by neutralizing reactive oxygen species that contribute to inflammation.

The inflammatory response is a complex process involving multiple genes and signaling pathways. By integrating the PPI network with the drug-active ingredient-target-disease network, we identified key genes—SRC, STAT3, MAPK1, HSP90AA1, EP300, CTNNB1, EGFR, and RELA—that exhibit high centrality in the network (degree ≥ 4 times the average value). These genes are hypothesized to be core targets mediating the anti-inflammatory effects of L. ruthenicum.

KEGG pathway enrichment analysis revealed that these core targets were primarily enriched in 20 signaling pathways, which can be categorized into three functional groups. The first group includes inflammation-related pathways such as neuroactive ligand-receptor interaction, calcium signaling, MAPK signaling, PI3K-Akt signaling, and Rap1 signaling pathways. The second group encompasses pathways related to infectious diseases, including human cytomegalovirus infection, Kaposi sarcoma-associated herpesvirus infection, COVID-19, and hepatitis B. The third group comprises cancer-associated pathways such as microRNAs in cancer, chemical carcinogenesis-receptor activation, proteoglycans in cancer, and chemical carcinogenesis-reactive oxygen species.

TNF-α and IL-1β are key regulators in various chronic inflammatory diseases. The activation of MAPKs triggers transcriptional cascades that promote the production of pro-inflammatory cytokines like TNF-α and IL-1β, which, in turn, activate a range of enzymes associated with the inflammatory response. Inhibiting MAPK kinases or blocking the overproduction of pro-inflammatory cytokines could mitigate chronic inflammatory diseases, as inflammation is closely linked to the development of numerous conditions. Based on these findings, we speculate that L. ruthenicum exerts its anti-inflammatory effects through the regulation of the MAPK signaling pathway.

The role of inflammation in cancer development has garnered significant attention in recent years. While inflammation generally serves as a protective response to injury or infection, prolonged or uncontrolled inflammation can increase the likelihood of genomic mutations and facilitate the overexpression of inflammatory mediators, creating an environment conducive to cancer development. Our analysis of the filtered signaling pathways indicates that cancer-related pathways are prominently involved in the pharmacological effects of L. ruthenicum. Although these pathways are not directly linked to inflammation regulation, they highlight the potential of L. ruthenicum in cancer therapy through its anti-inflammatory actions.

On one hand, these findings support the connection between inflammation and various diseases, suggesting that L. ruthenicum may be beneficial in the clinical treatment of cancer via its anti-inflammatory effects. On the other hand, our results also underscore the limitations of network pharmacology. While network pharmacology has become a powerful tool for studying the complex mechanisms of traditional Chinese medicine (TCM), it still has inherent limitations. Mathematical models and network simulations often fail to fully capture the intricacies of biological systems. Therefore, the reliability of data derived from network analyses requires further validation through pharmacological and pharmacodynamic experiments.