CR samples were prepared by reference methods (27). Fresh CR samples were vacuum freeze-dried in a freeze-dryer (Scientz-100F, Scientz, China) before being crushed to powder in a blender (30 Hz, 1.5 min, MM 400, Retsch, Germany). Approximately 50 mg of the sample was weighed, dissolved in 1.2 mL of 70% methanol (Merck, Germany) extract, and vortexed (IKA VORTEX GENIUS 3, VG 3 S25) every 30 min for 30 s for a total of 6 times. The sample was placed at 4°C overnight and centrifuged at 12,000 rpm for 10 min, and then, a 0.22-μm microporous membrane was used to filter the supernatant, which was then placed in a vial for an ultra performance liquid chromatography-electrospray ionization-tandem mass spectrometry (UPLC-ESI-MS/MS) analysis.

The CR extracts were analyzed by widely targeted metabolomics using UPLC-ESI-MS/MS (34). Chromatographic separation was done using an Agilent SB-C18 column (1.8 m, 2.1 mm, 100 mm; Agilent, Foster City, CA, USA) on a Nexera X2 UPLC machine (Shimadzu, Kyoto, Japan). Purified water with 0.1% formic acid (solvent A) and acetonitrile with 0.1% formic acid (solvent B) made up the mobile phase. The gradient elution program was as follows: 0–9 min (95% → 5% A), 9–10 min (5% A), 10–11.1 min (5% → 95% A), and 11.1–14 min (95% A). The flow rate was 0.35 mL/min, the injection volume was 2 μL, and the temperature in the column oven was set at 40°C. The effluent was alternatively introduced into a triple quadrupole-linear ion trap mass spectrometer (6500 Q-TRAP UPLC–MS/MS; Applied Biosystems, Framingham, MA, USA) equipped with an ESI Turbo Ion-Spray interface. The following settings were used to run the ESI ion source: ion source, turbo spray; temperature, 550°C; spray voltage, 5,500 V (positive ion mode)/-4,500 V (negative ion mode); gas I, 50 psi; gas II, 60 psi; curtain gas, 25.0 psi; and collision-activated dissociation, high. Instrument calibration and mass calibration were carried out with 10 and 100 mol/L polypropylene glycol solutions in triple quadrupole and linear ion trap modes, respectively. The multiple reaction monitoring (MRM) assays were used to collect triple quadrupole images with the collision gas (nitrogen) set to medium. The delustering potential and collision energy were optimized for individual MRM transitions to enhance their efficiency. Each period was examined for a specified set of MRM transitions based on the metabolites eluted during that interval. Analyst v1.6.3 (AB Sciex, Foster City, CA, USA) was used to collect and analyze mass spectra.

The resultant mass spectra were compared in the Metware Biotechnology Co., Ltd. (MWDB) database. Interference signals, such as the repeated signals of K⁺, Na⁺, NH⁴⁺ ions, the repetitive signals of fragment ions, and the isotope signal were initially eliminated throughout analyses to assure the correctness of the metabolite annotations. MRM-based quantification of metabolites was conducted using Multi Quant v3.0.2 (AB Sciex). After obtaining the signal intensities of each metabolite's distinctive ions in several samples, the chromatographic peak regions of those ions were integrated to indicate the metabolite's relative abundance. The peak areas were adjusted with retention time and peak type to make it easier to compare the same metabolites quantitatively between samples.

We used UPLC-ESI-MS/MS to analyze the metabolites to further understand the differences in metabolite accumulation of the five species of CR. Approximately 1,116 metabolites were found and identified in the five species of CR, divided into 14 different chemical classes (Figure 3C). The number of metabolites was 1,099 (LD) > 1,094 (WD) > 1,102 (CP) > 1,102 (CPM) > 1,093 (CT). Phenolic acids were the most prevalent class (214–216, 19.3%). The remaining classes were lipids (145–149, 13.3%), flavonoids (116–119, 10.5%), amino acids and their derivatives (117–119, 10.6%), organic acids (111–114, 10.1%), alkaloids (111–112, 10.0%), saccharides (76–78, 6.9%), nucleotides and derivatives (75, 6.7%), Others (33–35, 3.1%), lignans (34, 3.0%), coumarins (24–26, 2.2 %), vitamin (16–17, 1.5 %), terpenoids (10–13, 1.0 %), and quinones (2, 0.2 %) (Figure 3A).

Numerous investigations on the chemical composition of CR have been published thus far. More than 200 different substances had been identified from CR, including polynyas, polyenes, flavonoids, sugars, volatile oils, alkaloids, organic acids, and amino acids (8, 39). Approximately 1,116 compounds were discovered in this study that were based on an earlier study, considerably enhancing the compound database of CR and supplying the scientific basis for the CR quality control study. According to previous reports, phenylpropanoids, alkaloids, flavonoids, and polysaccharides make up the majority of CR's pharmacological constituents. These substances have pharmacological effects on the nervous, endocrine, and immune systems that manifest as anti-tumor, anti-inflammatory, anti-stress, and hepatoprotective effects (8). The metabolomic analysis revealed that the average number of phenylpropanoids in CR, comprising phenolic acids, lignans, and coumarins, was 274.2 (24.6%), accounting for the highest proportion. Phenolic acids are abundant in medicinal plants and have significant medicinal properties that can be made into medications or health foods with antibacterial, anti-inflammatory, analgesic, and anticancer properties. Among them, hydroxybenzoic acid and hydroxycinnamic acid are the main phenolic acids, which are derived from phenylalanine. They are important intermediates in the shikimic acid pathway and participate in the synthesis of subsequent lignans, flavonoids, stilbenes, and alkaloids (40). They are also aromatic secondary metabolites that give typical sensory properties to foods and are associated with a variety of health benefits (41). Our analysis also revealed that there were many kinds of alkaloids (111.6,10.0 %) and flavonoids (117.4,10.5 %) in CR. Modern pharmacological studies have shown that alkaloids and flavonoids have pharmacological effects, such as protecting the nervous system and promoting the circulatory system, in addition to anti-oxidation and anti-fatigue properties (8).

We also found 77.2 (6.9%) different species of saccharides in CR. Among different saccharides, polysaccharides are very significant macromolecules that are crucial for the development and growth of organisms (42). Studies on phytochemistry and biology have revealed that the primary bioactive components of CR are carbohydrates. The main element of CR that has immunomodulatory and anticancer actions is polysaccharide, a significant and comparatively non-toxic natural plant active biological macromolecule (43). At the same time, polysaccharides interact with flavor compounds in the food system, which has an effect on how flavor components are retained and released. CR polysaccharide is an important component affecting the flavor of CR (44). Further, we found high levels of lipids, organic acids, amino acids, vitamins, and other substances in CR. As the basic component of protein and the essential nutrients of the human body, amino acids not only determine the nutritional status of food but also affect the flavor of food and the physiological activity of drugs (45). Amino acids are important flavor substances. Aspartic acid and glutamic acid are umami, and alanine, proline, serine, threonine, and glycine are sweet (46). We found CR to be rich in amino acids (117.2,10.5 %) and a plant protein source with high medicinal value (47). In addition to rich nutrients, CR also has excellent flavor and food attributes related to taste. CR can be further developed and applied in food applications.

Approximately 1,054 metabolites were common to the five CR species, accounting for 94.44%, indicating that their metabolite accumulation characteristics were highly similar (Figure 3B). A total of 39 (3.49 %), 18 (1.61 %), and 5 (0.45 %) metabolites were detected in 4, 3, and 2 species of CR, respectively (Figure 3C). The five species of CR had no unique metabolites. This is because they were all related to the species, subspecies, or varieties of Codonopsis pilosula (Franch.) Nannf. in plant taxonomy, and their genetic relationship is very close. This supports the correctness in the Flora of China classification of C. pilosula, C. pilosula var. modesta, and C. tangshen as the species of Codonopsis pilosula (Franch.) Nannf. and the Chinese Pharmacopeia's classification of these three species as the source of CR. Meanwhile, LD and WD from Guizhou belong to the species included in the Chinese Pharmacopeia in terms of botanical taxonomy, and the chemical composition is highly similar to the traditional genuine producing areas. Therefore, LD and WD can be used and promoted as medicinal herbs. Compared to CP, CPM, and CT, 12, 14, and 18 metabolites were detected only in LD and 13, 10, and 18 metabolites were detected only in WD, respectively (Figure 3B). This might be related to the fact that Guizhou's production area is in a region with low longitude, low latitude, high altitude, heavy precipitation, high annual average temperatures, and average relative humidity. Moreover, both species (29, 48) and the growing environment have a strong influence on the accumulation of secondary metabolites in plants, and metabolite accumulation can vary to some extent between the same species based on their growing environments (27, 31).

The variation in the metabolic characteristics of medicinal plants is the result of the interaction of species, climate, altitude, rainfall, soil, and other environmental factors (32). Under relatively stable conditions, typical genuine medicinal material can be formed. When the habitat is changed or the species are artificially introduced to a different ecological environment, the accumulation of their metabolites will change accordingly; the greater the difference in environmental conditions, the greater their metabolic variation (49). The secondary metabolites of medicinal plants are the material basis of their potency and quality, and studies have confirmed that the quality of CR is significantly affected by species, origin, altitude, latitude, and longitude (50–52). Wu et al. (53) used the Max Ent model and Arc GIS technology to analyze the main environmental factors affecting the distribution of CR herbs by combining 55 environmental factors, and the results showed that precipitation and altitude had major effects on the distribution of all three CR species (C. pilosula, C. pilosula var. modesta, and C. tangshen). In conclusion, species and environmental variables affect CR quality, and these considerations should be completely considered when introducing artificial cultivation to prevent substantial financial losses.

To further evaluate the fundamental properties and variations of metabolites between LD and WD and CP, CPM, and CT, we carried out hierarchical cluster analysis and principal component analysis. The results of the cluster analysis revealed that the various CR samples were divided into two categories, where CP and CPM were clustered into one category and CT, LD, and WD were clustered into one category, indicating that CP and CPM metabolites had similar characteristics for accumulation, while CT, LD, and WD metabolites had similar characteristics for accumulation. In terms of metabolites, they were clustered into 6 categories, of which B, C, and E were dominated by phenolic components, accounting for 32%, 17%, and 20%, respectively. A, D, and F were dominated by lipids, amino acids and derivatives, and flavones, accounting for 33%, 28%, and 29%, respectively (Figure 4A). Principal component analysis showed that Dim1 and Dim2 accounted for 22.1% and 17.5% of the total variation, respectively. Dim1 significantly separated LD and WD (CP and CPM), and Dim2 significantly separated WD and CP, CPM, CT, and LD. Among the metabolites with a principal component score > 0.5, Dim1 was mainly composed of phenolic acids, organic acids, and lipids, while Dim2 was mainly composed of phenolic acids and lipids, indicating that Dim1 and Dim2 significantly affected phenolic acids, organic acids, and lipids (Figure 4B).

Differential metabolites were screened using | log₂(fold change) |>1 and VIP≥1 as criteria. LD and WD were compared with CP, CPM, and CT in a two-by-two comparison to obtain the following six comparison groups of differential metabolites: CP vs. LD, CPM vs. LD, CT vs. LD, CP vs. WD, CPM vs. WD, and CT vs. WD, whose specific expression of DMs is shown in Figures 6, 7. Compared with CP, CPM, and CT, 99, 68, and 105 metabolites were upregulated in LD, and 87, 67, and 135 metabolites were upregulated in WD. While LD upregulated metabolites were mainly concentrated in flavonoids, phenolic acids, alkaloids, and organic acids, WD upregulated metabolites were mainly concentrated in amino acids and derivatives, flavonoids, alkaloids, and organic acids (Figure 5). Previous studies showed that LD and WD had better food properties than CR from traditional production areas. It was speculated that the upregulation of the abovementioned DMs might be the reason based on our study. The comparative analysis found that there were 26, 9, 20, 16, 8, and 56 DMs upregulated only in CP vs. LD, CPM vs. LD, CT vs. LD, CP vs. WD, CPM vs. WD, and CT vs. WD, respectively. Including flavonoids (17 DMs), phenolic acids (3 DMs), amino acids and derivatives (2 DMs), alkaloids (1 DM), and coumarins (1 DM), 23 DMs were all upregulated in CP vs. LD, CPM vs. LD, and CT vs. LD. In the KEGG database, DMs were annotated to 7 metabolic pathways, mainly enriched to secondary metabolic biosynthesis (ko01110, 6 DMs) and flavonoid biosynthesis (ko00941, 6 DMs). Twenty-four DMs, including amino acids and derivatives (10), phenolic acids (7), alkaloids (4), others (1), and organic acids (1), were all upregulated in CP vs. WD, CPM vs. WD, and CT vs. WD. The KEGG database annotated 36 metabolic pathways, with the majority enriched to secondary metabolic biosynthesis (ko01110, 8 DMs) and metabolic pathways (ko01100, 13 DMs).

Compared with CP, CPM, and CT, 210, 301, and 180 metabolites were downregulated in LD, while 208, 268, and 215 metabolites were downregulated in WD (Figure 6). The comparative analysis found 19, 53, 27, 27, 25, and 33 DMs that were downregulated only in CP vs. LD, CPM vs. LD, CT vs. LD, CP vs. WD, CPM vs. WD, and CT vs. WD, respectively. Including phenolic acids (13 DMs), flavonoids (9 DMs), amino acids and derivatives (8 DMs), lipids (8 DMs), alkaloids (6 DMs), others (4 DMs), organic acids (3 DMs), lignans (2 DMs), and saccharides (1 DM), 54 DMs were all downregulated in CP vs. LD, CPM vs. LD, and CT vs. LD. In the KEGG database, DMs were annotated to 12 metabolic pathways, mainly enriched to metabolic pathways (ko01100, 5 DMs), tryptophan metabolism (ko00380, 3 DMs), and linoleic acid metabolism (ko00591, 3 DMs). Seventy-two DMs, including lipids (27 DMs), phenolic acids (14 DMs), flavonoids (8 DMs), others (7 DMs), coumarins (4 DMs), lignans (4 DMs), amino acids and derivatives (2 DMs), organic acids (2 DMs), alkaloids (2 DMs), saccharides (1 DM), and nucleotides and derivatives (1 DM) were all downregulated in CP vs. WD, CPM vs. WD, and CT vs. WD. The KEGG database annotated 15 metabolic pathways, with the majority enriched to secondary metabolic biosynthesis (ko01110, 9 DMs) and metabolic pathways (ko01100, 10 DMs).

Screening for metabolites with a p < 0.05 and their upregulation in LD or WD on the basis of metabolite difference analysis were required to determine the dominant differential metabolites of the two species of CR from Guizhou. From these two processes, 28 dominant differential metabolites were found in CP vs. LD, including yuanhuanin, isoorientin-7-O-(6^′′-feruloyl)glucoside, 4-Hydroxy-7-methoxycoumarin-β-rhamnoside, N-Phenylacetylglycine, and dioxindole-3-acetyl-3-O-glucoside; 25 prominent differential metabolites were detected in CP vs. LD, including 4,6-Dihydroxyquinoline, N-Phenylacetylglycine, 5,7-Dihydroxy-4-methylcoumarin, phlorizin, α-Hydroxycinnamic acid, and isoorientin-7-O-(6^′′-feruloyl)glucoside; and 33 prominent differential metabolites were discovered in CT vs. LD, including isoorientin-7-O-(6^′′-feruloyl)glucoside, 2-(Acetylamino)-3-phenyl-2-propenoic acid, N-Phenylacetylglycine, 3-Galloylshikimic acid, and phlorizin (Figure 7). A comprehensive investigation of three groups revealed that the three dominant differential metabolites, phlorizin, N-Phenylacetylglycine, and isoorientin-7-O-(6″-feruloyl) glucoside, were significantly upregulated in LD, with isoorientin-7-O-(6″-feruloyl) being detected in CR for the first time.

There were 57 dominant differential metabolites found in CP vs. WD, including dioxindole-3-acetyl-3-O-glucoside, N-Acetyl-L-Methionine, 7,8-Dihydroxy-4-phenylcoumarin, hesperetin-7-O-neohesperidoside(Neohesperidin), and alpha-hexylcinnamaldehyde. Approximately 25 prominent differential metabolites were detected in CPM vs. WD, including tryptamine, N-Acetyl-L-Methionine, 5,7-Dihydroxy-4-methylcoumarin, and abscisic acid. 2-(Acetylamino)-3-phenyl-2-propenoic acid^*, N-Acetyl-L-leucine, epigallocatechin, 3-Galloylshikimic acid, and sorbitol-6-phosphate were among the 76 prominent differential metabolites discovered in CT vs. WD (Figure 8). A comprehensive investigation of 3 groups revealed that the 10 dominant metabolites, L-Methionine, hydrocinnamic acid, radicamine A, 4-Hydroxymandelonitrile, L-Isoleucine^*, abscisic acid, N-Acetyl-L-Methionine, cinnamic acid, cyclo(Phe-Glu), and cyclo(Pro-Leu)^*, were significantly upregulated in WD, with N-Formyl-L-Methionine, 4-Hydroxy-L-Isoleucine, cyclo(Phe-Glu), cyclo(Pro-Leu)^*, cinnamic acid, hydrocinnamic acid, and 4-Hydroxymandelonitrile being among the seven prominent differential metabolites that were detected in CR for the first time.

By estimating the appropriate target proteins of the two dominant differential metabolites in LD, a total of 196 targets were found. The protein-protein interaction (PPI) network was constructed using the 196 targets and visualized using Cytoscape (3.7.0, Supplementary Figure S1A). The targets with ≥10 degrees were HSP90AA1, ESR1, MAPK1, HSPA8, LCK, CREBBP, HRAS, HSPA1A, APP, MAP2K1, PPARA, HSPA5, PARP1, and PARP1, suggesting a more crucial function for the targets in the PPI network map.

Using the DAVIDv database (6.8, https://david.ncifcrf.gov/), GO functional enrichment analysis, KEGG pathway enrichment analysis, and disease enrichment analysis were carried out on the 196 targets. For the statistically significant GO functional analysis, KEGG pathway enrichment analysis, and disease enrichment analysis, a p < 0.01 was chosen. Approximately 169 items, including 98 BP, 32 CC, 39 MF, 25 KEGG pathway analysis entries (Figure 9A), and 87 disease enrichment analysis entries (Figure 9B), were derived from the GO functional analysis (Supplementary Figure S1B). The top 20 entries of each were selected and visualized using the bioinformatics website (http://www.bioinformatics.com.cn/). Histone H3-K9 demethylation, notch receptor processing, protein phosphorylation, one-carbon metabolic process, amyloid precursor protein catabolic process, among others, were BP's primary inclusions in the GO functional analysis. Extracellular MF mostly comprises histone demethylase activity, dioxygenase activity, zinc ion binding, carbonate dehydratase activity, and hydro-lyase activity. CC primarily includes cytosol, plasma membrane, an important component of the plasma membrane, gamma-secretase complex, and other cellular components. Nitrogen metabolism, progesterone-mediated oocyte maturation, human T-cell leukemia virus 1 infection, prostate cancer, notch signaling pathway are among the major topics covered by KEGG pathway analysis. The illness enrichment analysis revealed that oncological diseases, cardiovascular system diseases, neurological disorders, metabolic system diseases, and metabolic diseases accounted for the majority of the 196 targets matching LD-dominating differential metabolites. A “drug-component-target-network” map of LD was created using Cytoscape software (3.7.0) based on the screening of 2 components, 196 targets, and 20 pathways (Supplementary Figure S2A).

The target proteins corresponding to the other nine prominent differential metabolites in WD were predicted, and a total of 487 targets acting on the nine active components were discovered since the Canonical SMILES of cyclo (Phe-Glu) could not be accessed from the PubChem database. The PPI network map was visualized using Cytoscape (3.7.0), and it revealed that the targets with ≥50 degrees were SRC, MAPK1, HSP90AA1, CTNNB1, HDAC1, GRB2, RXRA, CREBBP, RELA, ESR1, MAPK14, and PTPN11. This suggests that the targets play a more crucial role in the PPI network map (Supplementary Figure S2B).

GO functional enrichment, KEGG pathway enrichment, and disease enrichment were carried out on 487 targets using the DAVID database (6.8, https://david.ncifcrf.gov/). GO functional analysis, KEGG pathway analysis, and illness enrichment analysis were all chosen as statistically significant at a p < 0.01 each. Moreover, 645 GO functional analyses yielded a total of 645 entries (Supplementary Figure S3A), of which 407 were BP, 76 were CC, 162 were MF, 114 were KEGG pathway analyses (Figure 10A), and 361 were disease enrichment analyses (Figure 10B). The top 20 entries were chosen for bioinformatics visualization (http://www.bioinformatics.com.cn/). Chemical synaptic transmission, proteolysis, a reaction to lipopolysaccharide, an excitatory postsynaptic potential, a response to the medication, etc., were the primary components of BP in the GO functional analysis. CC mostly consists of a membrane raft, a plasma membrane, a presynaptic membrane, an integral component of the plasma membrane, and other membranes. The majority of the KEGG pathway analysis topics include neuroactive ligand-receptor interaction, cancer-related pathways, prostate cancer, apoptosis, calcium signaling pathways, etc. The 487 targets that corresponded to the major differential metabolites of WD were primarily associated with neurological and cardiovascular disorders, according to the disease enrichment analysis. The “drug-component-target-network” network map of WD was created using Cytoscape software (3.7.0) based on the nine components, 487 targets, and 20 pathways discovered during the screening (Supplementary Figure S3B).

Secondary metabolites from plants are a class of tiny molecules having a wide range of chemical structures and biological activity, as well as significant biological functions (29). In the plant world, flavonoids are extensively dispersed and have pharmacological actions, including anticancer, therapy for cardiovascular conditions, and hypoglycemia (40, 54, 55). Diabetes, cancer, bacterial infections, and neurological problems can all be treated with cinnamic acid and its phenolic acid metabolite derivatives (56, 57). An important family of chemical molecules known as alkaloids has a number of pharmacological properties, including analgesic, antibacterial, and anticancer properties (58, 59). The main differential metabolites of LD were flavonoids and amino acids, whereas those of WD were amino acids, alkaloids, phenolic acids, and organic acids, according to our screening of the prominent differential metabolites of LD and WD. By creating a “component-target-pathway” network topology, network pharmacology can graphically represent the efficacy of TCM through various components and targets, which is congruent with the holistic perspective and principles of diagnosis and treatment in TCM (60). We used network pharmacology to predict the pharmacological mechanisms of action of the dominant metabolites of LD and WD and discovered that the two dominant chemical components of LD may be closely related to nervous system protection, cardiovascular system improvement, oncological diseases, and diabetes treatment, while the nine dominant chemical components of WD may be closely related to nervous system protection and cardiovascular system improvement.