Two Chinese herbal medicines (Astragalus propinquus Schischkin, Pinellia pedatisecta Schott) were purchased from Xi’an International Medical Center Hospital (Xi’an, China). Astragalus propinquus Schischkin (batch number:1912003) and Pinellia pedatisecta Schott (batch number: 191112) were extracted separately. Astragalus propinquus Schischkin was extracted as follows: 0.5 kg of finely minced and dried roots of Astragalus propinquus Schischkin was extracted using 70% ethanol twice at 2 h each time in each experiment. Low-pressure drying methods were then used to process the extracts, which yielded 189 g of powdery extract. The powder was then dissolved in 70% ethanol and dried under low pressure to yield a total of 152.7 g of dry powder. The Astragalus propinquus Schischkin extract (AE) was stored at 4°C. Pinellia pedatisecta Schott was extracted as follows: 0.5 kg of dried stems of Pinellia pedatisecta Schott was powdered and then extracted using 70% ethanol for 2 h. The extract was filtered and lyophilized, which yielded 10.74 g of dry powder. Pinellia pedatisecta Schott extract (PE) was stored at 4°C.

The main components from A&P were characterized using HLPC-MS/MS analysis, configured with Waters BEH C18 Column (250 × 2.1 mm, 1.7 μm) maintained at 35°C. The mobile phase was constituted of 0.1% methanoic acid in water (A) and acetonitrile (B), using an optimal gradient program as previously established: 0–15 min, 95% (A)-5% (B); 15 min–25 min, 50% (A)-50% (B); 25 min–30 min, 5% (A)-95% (B); 30 min–35 min, 95% (A)-5% (B). The flow rate was adjusted to 0.3 ml/min, while analysis time of each component was 15 min. Mass spectrometry conditions: the positive and negative ionization modes of electrospray ionization (ESI) were used for detection. ESI Source conditions were as follows: Ion Source Gas1(Gas 1): 50, Ion Source Gas2(Gas 2): 50, Curtain Gas (CUR): 25, Source Tempreture: 500°C (positive ion) and 450°C (negative ion), Ion Sapary Voltage Floating (ISVF) 5500 V (positive ion) and 4400 V (negative ion), TOF MS scan range: 100–1,200 Da, product ion scan range: 50–1,000 Da, TOF MS scan accumulation time 0.2 s, product ion scan accumulation time 0.1 s. Second order mass spectrometry was obtained by Information Dependent Acquisition (IDA) and had a high-sensitivity mode. Declustering potential (DP): ±60 V, Collision Energy: 35 ± 15 eV.

The TCMSP database (http://tcmspw.com/) was used to retrieve pharmacological information of the compounds using “HuangQi” and “BanXia” as keywords. Two key ADME parameters, which were drug similarity (DL ≥ 0.18) and oral bioavailability (OB ≥ 30%), were then selected for the preliminary screening of the collected compounds (Pang et al., 2018). OB refers to the amount of a drug that actually enters circulation within the body after being ingested, while DL represents the similarity between the constituent and an established medication (Ru et al., 2014). A&P active compound targets were explored with the SwissTargetPrediction target prediction platform (http://www.swisstargetprediction.ch/).

The melanoma-related genes were extracted from three databases including DisGeNET Database (http://www.disgenet.org/), KEGG (https://www.kegg.jp/) and GeneCards Database (https://www.genecards.org/). The common target genes were screened out as the targets of A&P in the treatment of melanoma.

The A&P-related targets and the melanoma-related targets were intersected, and a Venn diagram of the cross-common targets was obtained. The Cytoscape software (v3.2.1) was used to create a PPI network of A&P with melanoma. To understand the complicated interactions between compounds and their corresponding common targets, the C-T network was constructed and displayed by utilizing the Cytoscape software (v3.2.1). The compounds and targets are represented by nodes and are linked with edges.

GO and KEGG functional enrichment analyses were carried out with the DAVID database (https://david.ncifcrf.gov/). Relevant pathway analyses with p values < 0.05 were interpreted to be significant and interesting. The Omicshare platform was used to visualize the results of GO and KEGG enrichment analyses. To visually elucidate the mechanisms of A&P in the treatment of melanoma, a network connecting components, core targets and pathways (C-T-P network) was established.

Among the compounds of A&P, the first 4 compounds with the higher concentration were used as ligands and their 2D structure were downloaded from PubChem (https://pubchem.ncbi.nlm.nih.gov/). Genes corresponding to these four compounds in the top eight core targets of PPI key protein “degree” value were selected as receptors and their 3D structures were obtained from the RCSB PDB database (https://www.rcsb.org/), including AKT1(PDB ID: 1UNQ), MAPK3 (PDB ID: 4QTB) and ESR1 (PDB ID: 7BAA). The protein was imported into AutoDockTools software for format conversion, and all water molecules and small molecule ligands of crystal molecules were removed, polar hydrogen was added, charge was loaded, and active pockets were constructed. AutoDock4 software was used for molecular docking with effective active ingredients, and the binding energy was less than “−5 kcal/mol,” it showed that the target had certain binding activity with the compound (Gaillard, 2018). Finally, Discovery Studio Visualizer software was used to analyze and visualize the 3D and 2D interaction diagrams.

Mouse melanoma B16F10 cells were procured from the American Type Culture Collection (ATCC, Manassas, VA, United States). The cells were cultured in high glucose Dulbecco’s modified Eagle medium (DMEM) containing 10% fetal bovine serum (HyClone, Logan, Utah, United States), 100 U/ml each of streptomycin and penicillin (Gibco, Grand Island, NY, United States) and were incubated at 37°C with 5% CO₂. The cells were then allowed to achieve 70–80% confluency before being exposed to 0.05% trypsin-EDTA and seeded onto 6- or 96-well plates (Corning, NY, United States).

A CCK-8 assay (Beyotime, Shanghai, China) was used to evaluate cell viability. The cells (1 × 10³ per well) were seeded onto 96-well plates and stored in DMEM until adherent to the flask wall. The cells were then exposed to either 1.0, 2.0, 3.0, 4.0, and 5.0 mg/ml AE or 0.2, 0.4, 0.6, 0.8, and 1.0 mg/ml PE for 24 and 48 h respectively. After the IC50 was calculated, AE and PE were combined into the wells in different concentrations and proportions for 24 h within the range of no cytotoxicity. At the end of each dose, 10 μl of CCK-8 was added and incubated for 4 h and the absorbance was recorded with a microplate reader at 450 nm (Epoch, Bio Tek, United States).

After a 24-h treatment period using various concentrations and A&P proportions, the concentration ratio of PE and AE was determined to be 1:2. To examine cell apoptosis, cultured B16F10 cells were exposed for 24 h to A&P, rinsed with phosphate-buffered saline, and treated with the Annexin V-FITC Apoptosis Staining/Detection Kit (Beyotime, Shanghai, China). Flow cytometry analysis was conducted with flow cytometry (BD, San Jose, CA, United States).

B16F10 cells were seeded in 6-well plates and incubated for 24 h at 37°C. Cells were treated with A&P (PE:AE = 1:2, the concentration of PE was 0.2 mg/ml) for 24 h. DMSO (0.1%) was used as the drug vehicle. For experimental groups, cells were exposed to A&P alone or in combination with 20 μM 740Y-P (PI3K/Akt pathway activator; MedChemExpress, China) for 24 h. Total RNA were extracted from B16F10 cells by the TRIzol reagent (Invitrogen, Carlsbad, CA, United States) in strict compliance with manufacturer instructions. RNA was reverse-converted into cDNA and subsequently amplified with the aid of a reverse transcription kit (RiboBio, Guangzhou, China) in accordance with the instructions. All primers are as follows: GAPDH (Forward: AAT​GGA​TTT​GGA​CGC​ATT​GGT, Reverse: TTT​GCA​CTG​GTA​CGT​GTT​GAT), PI3K (Forward: CGA​GAG​TGT​CGT​CAC​AGT​GTC, Reverse: TGT​TCG​CTT​CCA​CAA​ACA​CAG), Akt (Forward: CCC​TGC​TCC​TAG​TCC​ACC​A, Reverse: TGT​CTC​TGT​TTC​AGT​GGG​CTC), Bcl-2 (Forward: GAG​CCT​GTG​AGA​GAC​GTG​G, Reverse: CGA​GTC​TGT​GTA​TAG​CAA​TCC​CA) and Bax (Forward: AGA​CAG​GGG​CCT​TTT​TGC​TAC, Reverse: AAT​TCG​CCG​GAG​ACA​CTC​G). Quantitative Real-Time PCR (qRT-PCR) was carried out using the Applied Biosystems StepOnePlus Real-Time PCR System (Applied Biosystems, Singapore).

After treatment as above, total proteins were extracted from cells using a Cell Lysis Buffer (Beyotime, Shanghai, China). A BCA protein detection reagent (Beyotime, Shanghai, China) allowed for the evaluation of protein concentration. A 10% SDS-PAGE gel was used to separate the proteins into their components, followed by immunoblotting the proteins onto PVDF membranes. 5% milk in PBS plus 0.1% Tween 20 was used to block endogenous reactions for an hour. This was followed by overnight incubation of the membranes with the primary antibodies: Akt (#9272, 1:1,000), p-Akt (#13038, 1:1,000), Bcl-2 (#3498, 1:1,000) and GAPDH (#5174, 1:1,000) at 4°C. And then the membranes were incubated for 2 h at room temperature with HRP-conjugated secondary antibodies (Goat anti-rabbit IgG, 1:1,000). Finally, the bands were imaged using an enhanced chemiluminescence system (ECL, ZETA, Beijing, China). The protein band gray values were calculated using the ImageJ software. Each experiment was repeated three times.

A total of 16 male C57BL/6 mice (6–8 weeks old, 15.15 ± 0.29 g) were purchased from ChengDu Dossy Experimental Animals Co. Ltd (Chengdu, China). The mice were kept in a temperature-controlled chamber under a 12-h light-dark cycle with free access to food and water. Accurately 0.1 ml B16F10 cells (3 × 10⁶ cells per mouse) were subcutaneously injected into the right lower back region of mice. The mice weight and tumour length and width were measured every 2 days. The tumor volume was calculated using the following formula: tumor volume (mm³) = 1/2 × length × width² (mm³). When the tumor volume reached about 100 mm³, the mice were devided randomly into two groups: the A&P-treated group and control group (n = 4 per group). The A&P-treated group was intragastrically administered with A&P (25 mg/kg PE and 50 mg/kg AE) per day. Control group received the same dose of phosphate buffer saline. Data measurement was also taken per day. The mice were sacrificed by cervical dislocation after 15 days of treatment, and the tumors were harvested for the subsequent testing. All animal experiments were approved by the Ethics Committee of Xi’an International Medical Center Hospital.

Frozen section samples of tumor tissues were prepared, cut into 5 μm-thick and then were immersed in 4% paraformaldehyde. For TUNEL staining, sections were digested with proteinase K for 25 min at 37°C and then incubated with the TUNEL reaction mixture (Roche) for 1 h in a dark-wet box at 37°C. For Ki-67 staining, the slices were immersed in 0.01 M citrate buffer (PH 6.0) and placed in the microwave oven on high heat until boiling and power off, repeated after an interval of 5 min. After washed two times with PBS, samples were incubated with Ki-67 antibody (abcam; 1:250) overnight at 4°C and then incubated with secondary antibody at 37°C for 30 min. Finally, the sections were observed and photographed under a microscope (Motic).

GraphPad Prism 6.02 (San Diego, CA, United States) was used to perform all statistical analyses, will all findings depicted in terms of mean ± SD. The Student’s t-test or two-way analysis of variance (ANOVA) was used to analyze data. p < 0.05 was considered to be significant.

HPLC-MS/MS was adopted to identified the main medicinal components in A&P. According to the chromatograms, a total of nine chemical constituents were identified (Figure 2). The concentrations of the compounds were listed in Supplementary Table S1. In addition to them, bifendate was hardly detected by HPLC-MS/MS, but it had 27 common targets, the largest number, with melanoma-related pharmacological targets. And it had an anti-cancer effect against multidrug-resistant tumors (Ren et al., 2016). So bifendate was also considered as a candidate component. To retrieve the active components of A&P, two key ADME parameters, DL (≥0.18) and OB (≥30%) were selected for the preliminary screening of the collected compounds. Although some ingredients do not meet the screening criteria, they have clinical efficacy; We also retained these for a full analysis of the components that we studied. For example, OB of rutin was very low and coumarin has a low DL, but they are the major components of Astragalus propinquus Schischkin, which had cytotoxic effects in melanoma cells (Bhattarai et al., 2021; Pinzaru et al., 2021). Another component, protocatechualdehyde, an active constituent from Pinellia pedatisecta Schott, has a low DL, but it induces cell apoptosis in melanoma B16-F10 cells (Zhong et al., 2020). Therefore, they were also included as candidate components. Hence, a total of 13 ingredients were selected as active ingredients in A&P as shown in Table 1. Specific information of these components are provided in Supplementary Table S2.

A total of 487 pathogenic targets were found to be associated with the compounds of A&P from the TCMSP (Supplementary Table S3). 464 melanoma-related pharmacological targets were screened out from DisGeNET Database and KEGG Platform (Supplementary Table S4). 70 common targets were determined to function as potential A&P targets in melanoma treatment (Figure 3A). Specific information of the 70 common targets are described in Supplementary Table S5. Data from 70 common targets were incorporated into STRING to build a PPI network to explore the interactive relationships with each other. And then, the PPI network was optimized by Cytoscape v3.2.1 (Figure 3B). In the PPI network, the nodes with higher degree might be more important in melanoma treatment. The detailed information of the PPI network are presented in Supplementary Table S6. 26 core targets with degree values above the median were identified. Among them, the top eight targets were tumor protein p53 (TP53), heat shock protein 90 alpha family class A member 1 (HSP90AA1), AKT serine/threonine kinase 1 (AKT1), mitogen-activated protein kinase 3 (MAPK3), E1A binding protein p300 (EP300), estrogen receptor 1 (ESR1), mitogen-activated protein kinase 1 (MAPK1), phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA).

Based on the 13 compounds and 70 intersection targets, we constructed a C-T network by Cytoscape v3.2.1. As shown in Figure 3C, 13 hexagonal nodes represent the active compounds of A&P, and 70 triangular nodes represent the common targets of A&P in melanoma treatment. The network showed the interrelation between the compounds of A&P and the target genes. The detailed information of the network are displayed in Supplementary Table S7. The compounds with higher degree of connections were bifendate, baicalein, isorhamnetin and quercetin. For the target analysis, ESR1 was linked to the largest number compounds. In addition, AKT1, HSP90AA1 and MAPK3 were respectively connected to four compounds. These findings suggest that A&P treats melanoma with multiple compounds acting on multiple genes.

GO enrichment analysis of the 70 potential therapeutic targets was carried out to characterize the candidate biological functions of A&P in treating melanoma, including biological process (BP), cellular component (CC), and molecular function (MF) categories. A total of 408 BPs, 38 CCs, 85 MFs and 152 pathways (p < 0.05) were obtained from the DAVID database (Supplementary Table S8). The BP analysis showed the key targets to be significantly correlated with positive regulation of gene expression, negative regulation of apoptotic process, positive regulation of transcription from RNA polymerase II promoter and protein phosphorylation. The CC analysis were associated with macromolecular complex, nucleoplasm, nucleus and cytoplasm. The MP analysis suggested the main targets to be related to protein serine/threonine/tyrosine kinase activity, enzyme binding, ATP binding and protein kinase activity. The top 10 BPs, CCs, and MFs items are shown in Figure 4A. The top 20 potential pathways of KEGG analysis results are presented in Figure 4B. Among them, there were eight signaling pathways, including PI3K-Akt signaling pathway, C-type lectin receptor signaling pathway, Cellular senescence, FoxO signaling pathway, PD-L1 expression and PD-1 checkpoint pathway in cancer, Neurotrophin signaling pathway, EGFR tyrosine kinase inhibitor resistance and apoptosis. The analysis results showed that the PI3K-Akt signaling pathway was the top signaling pathway. This pathway contains many key genes of 26 core targets, including TP53, HSP90AA1, AKT1, EGFR, MAPK1, MAPK3, SYK, IL2, MTOR, NFKB1, IL6, PIK3CA, CDK2, BCL2 and MDM2. These results suggest that the PI3K-Akt signaling pathway may be a vital pathway for A&P treating melanoma.

We proceeded to establish a C-T-P network constructed from 26 core targets, 13 active compounds and 8 signalling pathways, in order to characterize the therapeutic relationship between A&P and melanoma (Figure 5). Several different targets within these pathways were appeared to be associated with melanoma treatment. This network depicts direct relationships between compounds, targets, and pathways. Purple hexagonal nodes indicate compounds, red angular nodes indicate pathways, yellow round nodes indicate targets, and lines indicate the interactions between them. Additionally, the PI3K-Akt signaling pathway covered the most targets. In addition, we further isolated the network relationships for additional annotation of the KEGG pathway. The network places significant emphasis on the PI3K-Akt signaling pathways (Figure 6). This suggests that A&P may act on melanoma by modulating the PI3K/Akt signaling pathway.

According to the above results, we selected four compounds of higher content in A&P and their corresponding core targets to do molecular docking, including Quercetin -AKT1, Hederagenin-MAPK3, Beta-sitosterol-MAPK3, Beta-sitosterol-ESR1, Stigmasterol-MAPK3 and Stigmasterol-ESR1. The docking results as follows: Quercetin was docked to AKT1 with four hydrogen bonds LYS14 (2.1 Å), TYR18 (2.4 Å), ARG23 (2.4 Å) and ASN53 (1.8 Å) and by forming hydrophobic interactions with TYR18 (Figure 7A). Hederagenin could bind to MAPK3 by forming hydrophobic interactions with eight residues (ILE48, ALA69, TYR53, VAL56, LYS71, LEU124, LEU173, CYS183) and four hydrogen bonds with GLN122 (2.1 Å), LYS168 (3.1 Å), SER170 (2.2 Å) and ASN171(2.0 Å) (Figure 7B). Beta-sitosterol occupied the hydrophobic pocket composed of the residues (TYR53, VAL56, ALA69, LYS71, ILE73, ILE101, ILE120, LEU173, CYS183) and form a hydrogen bond with ASP123 (1.8 Å) in MAPK3 (Figure 7C), as well as the hydrophobic pocket composed of the residues VAL46, PHE119, PRO167, ILE168 and form a hydrogen bond with GLU115 (1.9 Å) in ESR1 (Figure 7D). Stigmasterol was docked to MAPK3 mainly by hydrophobic interactions with 10 residues (ILE48, TYR53, VAL56, ALA69, LYS71, ILE73, ARG84, ILE101, LEU173, CYS183) (Figure 7E). Stigmasterol bound to ESR1 by forming hydrophobic interactions with seven residues (CYS42, LYS49, VAL46, LYS122, PHE119, PRO167, VAL595) and a hydrogen bond with ASP215 (1.8 Å) (Figure 7F). The results showed that these receptors had strong bindings with the ligands (Table 2). Based on that AKT1, MAPK3 and ESR1 were included in cancer pathways, as well as AKT1 and MAPK3 included in the PI3K-Akt signaling pathway, the docking results again confirmed the prediction of network pharmacology.

Firstly, different concentrations of AE (0–5.0 mg/ml) and PE (0–1.0 mg/ml) were used to treat B16F10 melanoma cells. Both AE and PE remarkably inhibited malignant cell growth in a dose-dependent manner (Figure 8A). Furthermore, the half-maximal inhibitory concentration (IC50) of cells was determined at 24 h. The IC50 of AE on B16F10 melanoma cells was 2.86 ± 0.12 mg/ml, and the IC50 of PE on B16F10 melanoma cells was 0.65 ± 0.23 mg/ml. Based on the IC50 value, compounds PE and AE acted on cells at ratios of 1:1, 1:2, 1:5, and 1:10 (Supplementary Figure S1). The results showed that compound PE and AE at a concentration of 1:2 blocked malignant cell growth in a dose-dependent manner (Figure 8B). Cell apoptosis increased after compound PE and AE treatment. Flow cytometry analysis demonstrated that the early and late apoptosis of B16F10 cells was significantly promoted (Figure 8C) by compound PE and AE treatment.

A&P appears to exert its therapeutic effect on melanoma through 26 active compounds targeting 62 melanoma-associated genes that influence the PI3K/Akt and apoptosis signaling pathways. Cell apoptosis was found to be enhanced by A&P, and we sought to further investigate whether the anti-tumor effects of A&P on B18-F10 cells were mediated by inhibition of the PI3K/Akt pathway. B16F10 cells were treated with A&P (PE:AE = 1:2, the concentration of PE was 0.2 mg/ml) alone or in combination with 20 μM 740Y-P. qRT-PCR analysis demonstrated that A&P suppressed the relative expressions of PI3K, Akt, and Bcl-2 mRNA and promoted the relative expression of Bax mRNA and the addition of 740Y-P could reverse the changes by A&P (Figure 9A). Meanwhile, western blot analysis revealed that the expression of p-Akt and Bcl-2 were reduced in A&P-treated group, but was rised in the A&P + 740Y-P group (Figure 9B). As a whole, these data supported the hypothesis that A&P stimulated B16F10 cell apoptosis by acting on the PI3K/Akt pathway.

As shown in Figure 10A, In vivo experiments the B16F10 cell xenograft model was established to further investigate whether A&P could effectively inhibit tumor growth. After 15 days of treatment, the mice were sacrificed. The results showed that the tumor volumes from the A&P-treated group were remarkly smaller in comparison with that in the control group (Figure 10B), while the bodyweights of mice were no significant differences between these two groups (Figure 10C). TUNEL and Ki-67 staining results showed that more apoptic cells and fewer proliferative cells were observed in the A&P-treated group than that in the control group (Figure 10D). These findings demonstrated that A&P suppressed the B16F10 tumor growth in vivo.