Every 100 g herb (AM and RT, from Beijing Tongrentang) was soaked in deionized water for 1 h and decocted twice, separately. The decoction was poured into a beaker for 50 ml concentration, transferred, and centrifuged at 1000 g for 20 min. The supernatants, respectively, were the Astragalus Membranaceus decoction (AMD) and Radix Trichosanthis decoction (RTD).

The LC-ESI-MS/MS analysis system was equipped with an ultrahigh-performance liquid chromatograph (UHPLC) and an MS system. A Waters ACQUITY UPLC HSS T3 C18 column (2.1 mm × 100 mm, 1.8 μm) was used at 30°C and 5 ul of injection volume. Gradient elution conditions were set as given in Supplementary Table S1. MS parameters were a sheath gas flow of 40 arb, an auxiliary gas flow rate of 15 arb, a capillary temperature of 320°C, an auxiliary gas heater temperature of 350°C, and a positive spray voltage of 3.2 kv. The resolution of MS and MS/MS was 70,000 and 17,500, respectively.

The research process is shown in Supplementary Figure S1. The active chemical compositions of AM and RT were retrieved from the Traditional Chinese Medicine System Pharmacology (TCMSP) database and screened, which conformed to the oral bioavailability (OB) ≥ 30% and drug-likeness (D)L ≥ 0.18. The potential targets of herbs were further screened in target relative databases and completely reserved. The known potential melanoma-related targets were found with “Homo sapiens” in disease target databases and merged and deleted replicas of all targets. We used the STRING database to establish a protein–protein interaction (PPI) network and visualized the PPI by Cytoscape software. Subsequently, KEGG analysis was performed by linking targets to the Database for Annotation, Visualization, and Integrated Discovery database (DAVID). The drug–target–pathway–disease network was established via Cytoscape software.

Human malignant melanoma A375 cells and human L-02 liver cells (purchased from Cell Resource Center, Peking Union Medical College) were plated onto 96-well plates with 3×10³ cells/well. A control group (medium), AMD groups (medium with concentrations 2, 4, 8, 10, and 20 mg/ml of AMD), and RTD groups (medium with concentrations 0.16, 0.32, 0.64, 1.28, and 2.56 mg/ml of RTD) were set and incubated for 48 h or 72 h. After MTT solution and DMSO were used successively, the OD value was estimated at an absorbance of 490 nm. IC₀, IC₂₅, and IC₅₀ (the cell inhibitory rates were 0, 25, and 50%, respectively) were calculated. Afterward, the control group, different concentrations of AMD, RTD, and “cocktail” groups were set up for further tests.

Male and female C57BL/6J mice aged 6–8 weeks (purchased from Huafukang Biotechnology) were randomly divided into six groups of six with the abdomen depilated. Following the experimental progress (Figure 4A), 0.1 ml PBS was subcutaneously injected at the abdomen of the mice of the normal control group (NCG), while suspension (prepared by 0.1 ml of PBS) with 5 × 10⁵ B16-F10 cells were injected into others. This day was marked as d0. On d2, small black spots greater than 2 mm diameter appeared at the injection site. At d3, the NCG and model control group (MCG) were given ddH₂O intragastric administration, while the three compatible groups were given AMD and RTD solution at a ratio of 5:3 (L 3.12 g/kg, M 6.24 g/kg, and H 12.48 g/kg, respectively), and the positive control group (PCG) was given compound cyclophosphamide solution (19.5 mg/kg) once a day for 14 consecutive days. The mice weight, tumor length, and transverse diameter were recorded every second day, and the tumors of MCG were photographed every 4 days. Finally, the mice were anesthetized, and the blood, tumor, and liver were collected. The pathological features were determined by H&E staining. All these experiments were in accordance with the guidelines of the Animal Care and Use Committee.

After treating A375 cells with AMD, RTD, and their “cocktail” groups, the cell cycle and apoptosis (processed by an Annexin-FITC Apoptosis Detection Kit) were observed by flow cytometry. In addition, we stained the cells with Hoechst 33342 and photographed them under a confocal microscope (400x image) to observe the apoptotic morphology.

After A375 cells were cultured for 24 h, a cross was scraped at the center of wells by using a 10-µl micropipette tip and photographed to record. Then, the cells were cultured after they were treated with different concentrations of AMD, RTD, and their “cocktail.” After 48-h or 72-h treatment, images were captured again.

The treatment of A375 cells was consistent as mentioned before; then, the cell supernatant (treated for 48 h) and mice serum were collected (treated for 14 days). As per the guidelines of the corresponding kit manufacturer’s instructions, the value of LDH, SOD, and MDA was measured. Finally, the results were calculated by using the formula given in the instructions.

The protein in tissues and cells were extracted to determine protein concentration by using a BCA protein detection kit. Western blot experiments for β-actin, Bcl-2, Bax, p53, E-cadherin, N-cadherin, snail, Akt, and p-Akt were performed and photographed by using the Gel imaging instrument. ImageJ software was used to analyze the bands.

The experiments were repeated at least three times, and differences between all data were evaluated using ANOVA, and p < 0.05 was considered statistically significant.

The Compound Discoverer 3.1.0.305, mzCloud, and mzVault databases were used to identify unknown compounds. The chemical components are listed in Supplementary Table S2, S3 and Table 1. There were 25 and 49 chemical components of AMD and 14 and 16 components of RTD in the negative and positive ion modes, respectively (Figure 1), while 20 and 47 components were observed in their cocktail at a ratio of 5:3 in different ion modes (Table 1). The AMD compounds include flavonoids, organic acids, and amino acids. Terpenoids, organic acids, and amino acids are the main compounds in RTD. Compounds of the cocktail of AMD and RTD are also flavonoids, organic acids, and amino acids.Interestingly, cocktail contains some unique compounds like berberine and diosmetin-7-O-β- Dglucopyranoside.

Overall, 20 active chemical compositions of AM and 2 of RT were obtained via the TCMSP database (Supplementary Table S4). According to aforementioned databases, there are a total of 691 component targets of herbs, and 901 known potential melanoma-related targets were obtained. After summarizing these targets, 107 common targets were obtained, as shown in the Venn diagram in Figure 2A. According to the degree of alignment, the key proteins included SRC, PIK3CA, PIK3R1, Akt1, MAPK1, HRAS, RAC1, MAPK3, HSP90AA1, and PTPN11 (Figure 2B). According to KEGG analysis, 34 signaling pathways were obtained (Figure 2C). The action pathways mainly include PI3K-AKT, FOXO, and VEGF signaling pathways. The drug–target–pathway–disease network was established to reveal the relationship between them (Figure 2D).

A375 cell proliferation was significantly inhibited by AMD and RTD in a dose-dependent manner (Figure 3A), while we found there was a low inhibitory effect on L-02 cells (Figure 3B). The corresponding IC₀, IC₂₅, and IC₅₀ of each period were calculated (Table 2). They were used to further assess the different effects on A375 cell proliferation with AMD, RTD, and their “cocktail.” The proliferation inhibition rate of “cocktail” groups was higher than those of the corresponding AMD and RTD groups (p < 0.05) (Figure 3C). Among all the “cocktail” groups, AMD IC₂₅ + RTD IC₅₀ (compatibility in a ratio of 5:3) had a stronger inhibition than others on the A375 cell proliferation (Figure 3C).

We had already confirmed that the compatibility group has better effects than the single drug, and the compatibility has an additive effect. Therefore, in the animal experiment, we directly selected the best compatibility ratio (5:3) to detect its effect on mice with malignant melanoma. After B16-F10 cells were injected 2 days later, the tumor began to form—2 mm in diameter (Figure 4B). Then the mice body weight in most groups showed an increasing trend, while the PCG obviously decreased compared with NCG (Figure 4C), and the spirit and activity of the mice had no obvious change. Compared with MCG, the tumor weight of medication administration groups was decreased significantly (p < 0.05) (Figure 4D). The H&E staining results indicated a large area of tumor tissues had disappeared in administration groups (Figure 4E), compared with the MCG. Interestingly, there was no significant difference in the pathological sections of liver tissues in all groups.

To investigate the mechanism of tumor growth and cell proliferation inhibition, we further analyzed the cell cycle and apoptosis in vitro. AMD, RTD, and their “cocktail” could play an inhibitory role in cell proliferation through the induction of cell cycle arrest (Figure 5). Furthermore, AMD, RTD, and their “cocktail” could promote A375 cell apoptosis in a dose-dependent manner, and the numbers of early and total apoptotic cells increased significantly in the “cocktail” groups, which showed an additive effect of compatibility (Figure 6A). Hoechst staining results showed that the most nuclear membranes in groups RTD IC₅₀ (AMD IC₀+RTD IC₅₀) and (AMD IC₂₅ + RTD IC₅₀) were obviously wrinkled, and the chromatin in nuclei was condensed (Figure 6B). Meanwhile, the herbs could reduce Bcl-2 expression and enhance the Bax protein level, with a ratio of Bax/Bcl-2 significantly increased in a dose-dependent manner (Figures 6C,D), and the “cocktail” groups had better effects, especially (AMD IC₀+RTD IC₅₀) and (AMD IC₂₅ + RTD IC₅₀). In addition, we found that the herbs could also increase p53 protein expression, and group (AMD IC₂₅ + RTD IC₅₀) had the best effect (Figure 6E). In vivo, the compatibility treated in the melanoma model could also promote p53 and Bax protein expression (Figure 6F).

The wound healing assay showed that the control group suffered a closing tendency of a scratch wound and the cell number increased, while the others led to slower closing of a scratch wound with statistical difference (Figure 7A). Thus, AMD, RTD, and their “cocktail” could inhibit the A375 cell migration in a dose-dependent manner. The “cocktail” groups had a stronger cell migration inhibition, especially groups (AMD IC₀+RTD IC₅₀) and (AMD IC₂₅ + RTD IC₅₀).

Furthermore, in treatment groups of A375 cells, especially “cocktail” groups (AMD IC₀ + RTD IC₅₀) and (AMD IC₂₅ + RTD IC₅₀), E-cadherin increased and N-cadherin decreased, resulting in an increased ratio of E-cadherin/N-cadherin (Figures 7B,C). In addition, snail expression of A375 cells decreased (Figure 7D). In vivo, E-cadherin increased and N-cadherin decreased in tumor tissues treated by the compatibility (Figure 7E).

LDH is a key enzyme involved in tumor glycolysis and associated with tumor proliferation and metastasis. SOD and MDA are markers of oxidation and antioxidation and are also related to tumor apoptosis and metastasis. Therefore, we examined their expression. After treatment with AMD, RTD, and their “cocktail” for 48 h in A375 cells, the LDH activity of herb-only groups decreased, and the “cocktail” groups (AMD IC₂₅ + RTD IC₀) and (AMD IC₂₅ + RTD IC₅₀) had an additive effect with a statistical difference (Figure 8A). In vivo, compared with the NCG, the LDH activity of the MCG significantly increased, and that in other groups decreased (Figure 8B). In addition, the MDA content increased and the SOD activity decreased, in a dose-dependent manner. The “cocktail” groups, especially (AMD IC₀+RTD IC₅₀) and (AMD IC₂₅ + RTD IC₅₀), had an additive effect (Figure 8C, D).

In vitro, we found that after herb treatment for 48 h, certain doses of AMD and RTD led to a decrease in Akt and p-Akt expression. For 72 h, herbs, especially “cocktail” groups (AMD IC₀+RTD IC₅₀) and (AMD IC₂₅ + RTD IC₅₀), could reduce the expression of Akt and p-Akt. In vivo, their compatibility also could reduce the Akt protein expression in melanoma tissues (Figure 9D).