The TanIIA-targeted genes were obtained from TCMSP (https://tcmsp-e.com/) (17), Herb (http://herb.ac.cn/) (18), PubChem (http://pubchem.ncbi.nlm.nih.gov/), STITCH databases (http://stitch.embl.de/) (19) and HIT (http://hit2.badd-cao.net) (20) by searching “Tanshinone IIA”. In addition, the SMILE format of TanIIA was copied to SwissTargetPrediction database (http://www.swisstargetprediction.ch/) (21) to predict the relevant targets of Probability > 0.1. Moreover, the TanIIA-targets and gene names were screened by Normfit ≥ 0.3 and Zscore≥ 0.5 from PharmMapper database (http://lilab.ecust.edu.cn/pharmmapper/ ) (22) via uploading the 3D structure of TanIIA. The final targets were used to construct a TanIIA-TanIIA target network.

HCC-related genes were obtained from the GeneCards database (https://www.genecards.org), OMIM (https://www.omim.org) and Drugbank databases (https://www.drugbank.com/) using “hepatocellular carcinoma” or “hepatoma” as keywords. Additional related genes were collected from the OncoDB.HCC database(http://oncodb.hcc.ibms.sinica.edu.tw/index.htm) (23).

To obtain the potential targets for TanIIA treatment of HCC, the HCC-related genes were crossed with the TanIIA-targets. Then the intersection targets were introduced into the STRING database (https://www.string-db.org/). The targets with confidence score > 0.4 were employed to carry out the protein-protein interaction (PPI) network visualization and analysis. The degree of each protein was calculated by Cytoscape3.8.0 software plug-in, and the proteins with a degree value > 5 were screened as TanIIA-related genes.

The R package “clusterProfiler” was employed to perform the functional enrichment analysis of the intersection targets, including gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses. Terms and pathways with P values < 0.05 were screened, and the top 15 enriched GO terms and KEGG pathways were visualized via “ggplot2” package (24).

The Univariate Cox analysis conducted via the “survival” R package was employed to find out the TanIIA-related prognostic genes (p< 0.05). And the efficient prognostic model was established by LASSO analysis with “glmnet” R package. And the overall survival (OS) was chosen as the response variable. Meanwhile, the minimal penalty term (λ) with ten-fold cross-validation was unitized. And the risk score of HCC patients from the TCGA cohort was calculated according to the formula: risk score=sum (gene expression level × corresponding coefficient). Thereafter, the patients were classified into low-risk and high-risk groups on basis of the median risk score. The “survminer” and “timeROC” package were used to conduct survival analysis and predictive accuracy.

The GEPIA2 database (http://gepia2.cancer-pku.cn/) (25) was employed to examine the expression of the prognostic genes in HCC. The Spearman’s correlation analysis and “pheatmap” package were performed to analyze and display the correlation between prognostic genes. The genes which had statistically significant correlation with the candidates were considered as the core genes.

The chemotherapeutic sensitivities to sorafenib between difference expression of the prognostic genes were predicted by the R package “pRRophetic”. In accordance with the Genomics of Drug Sensitivity in Cancer (GDSC) database (https://www.cancerrxgene.org/), the IC₅₀ estimates were obtained via ridge regression.

Single-sample gene set enrichment analysis (ssGSEA) was applied to explore the relationship between the expression of core genes and immune cells, and the infiltrations of immune cells were quantified utilizing the GSVA package. In addition, the “ESTIMATE” package was employed to compute the immune, ESTIMATE, and stromal scores. Then, the correlation between the expression levels of core genes and immunity were analyzed by the Spearman’s correlation.

Molecular docking models for the identified core genes and TanIIA were constructed to evaluate prognostic targets. The PDB formats of the above compounds were retrieved from RCSB (http://www.rcsb.org) database. 3D chemical structural formulas were made via ChemBioDraw 3D. Then they were modified by the SYBYL-X (version 2.0, TRIPOS Inc.) software, including ligands and water removal, amino acid patching and optimization, hydrogen addition and active pocket construction. Moreover, SYBYL-X software was used for docking and evaluating the docking scores, which represents of binding affinities between TanIIA and core genes. Finally, the results were also visualized in SYBYL-X software. In general, the CSCORE >4.0 indicates good binding activity with the molecules (26).

The two hepatocellular carcinoma cell lines, HepG2 and Hep3B, were purchased from Procell Life Science & Technology Co. Ltd. (Wuhan, China) and cultured in DMEM medium with 1% streptomycin and penicillin (KeyGEN biotech, Nanjing, China) supplemented with 10% FBS (GIBCO, NY, US). The cells were incubated at 37°C with 5% CO₂.

TanIIA (purity ≥98%) was purchased from Yuanye Biotech (Shanghai, China) and dissolved in DMSO at concentrations of 5 m. The TanIIA solution was stored at -20°C.

Cell proliferation was determined by the CCK-8 assay (Beyotime Institute of Biotechnology, Shanghai, China). HepG2 and Hep3B cells were seeded in 96-well plates at a density of 3 × 10 (4)/well. When adhered to the wall, the cells were intervened with various concentrations of TanIIA at 0, 20, 40, 60, 80, 100 and 120 μmol for 24h. The control (non-intervened) cells (0 μmol) were in DMEM containing 0.3% DMSO. According to the manufacturer’s protocol, CCK-8 reagent (10 µL/well) was added to the cells, and cells were cultured for 30min at 37°C. The absorbance was measured by a microplate spectrophotometer (Thermo Scientific, USA) at 450 nm. The cell viability rates (%) of the two cells were calculated as experimental absorbance value/control absorbance value × 100. Moreover, The IC₅₀ were evaluated via GraphPad Prism 8.0 software.

The detection of TanIIA-induced apoptosis was determined by Flow cytometry assay (FCM). Hep3B and HepG2 cells were respectively implanted in 6-well plates at a density of 1 × 10 (5) cells/well overnight. Then, the cells were intervened with a series of different concentrations of TanIIA solution for 24 h. After harvested, the above cells were washed and resuspended gently with PBS. Whereafter, 5 μL Annexin V-FITC and 5 μL PI staining solution were added into cells. They were mixed gently and incubated in the dark at room temperature for 15 min. The cell apoptosis was then detected by FCM (BD Biosciences, USA).

Transwell assay was employed to measure the invasive ability. Hep3B and HepG2 cells were intervened with TanIIA for 24 h. The inserts (8-mm pore size; corning, inc.) in the 24-well plate were coated with diluted Matrigel (BD Biosciences, USA) initially. The intervened cells were trypsinized and resuspended in serum-free medium. Subsequently, 200μl cells (1×10 (5)) were transferred in the upper chambers, and the 24-well plate was filled with 500 μL 10% FBS-containing complete medium. Following 24 h of reseeding, the cells were subjected to 30min 4% paraformaldehyde fixation and 20min 0.1% crystal violet staining. After washing with PBS several times, the cells attached to the upper chamber were removed gently and the inferior surface were visualized under a ×10 microscope. Finally, the cells were quantified utilizing ImageJ software.

The TanIIA-intervened cells were lysed in RIPA buffer (Vazyme Biotech, China). Afterward, 30 µg proteins were isolated by 10% SDS-PAGE and transferred to PVDF membranes (0.45μm). Afterwards, the membranes were blocked in 1-hour 5% non-fat milk at ambient temperature. Subsequently, these blots were incubated overnight at 4°C with specific primary rabbit antibodies against ALB (Proteintech, China, diluted 1:1000), SRC (Proteintech, China, diluted 1:300), ESR1 (Proteintech, China, diluted 1:500), BAX (Cell Signaling Technology (CST), USA, diluted 1:1000), BCL-2 (CST, USA, diluted 1:1000), GAPDH (Proteintech, China, diluted 1:2000). The next day, the membranes were incubated with goat anti-rabbit secondary antibody (Sangon Biotech, China, diluted 1:5000) for 1 hour. In the end, the chemiluminescence image analysis system (Tanon, Shanghai, China) was utilized for protein detection and visualization. The proteins were quantified by ImageJ software, and the relative levels were normalized by comparison to GAPDH.

The 3D structure of TanIIA is shown in Figure 2A . Its pharmacological and molecular properties, obtained from TCMSP, are shown in Supplementary Table 1. The number of targets searched from 7 databases was as follows: there were 40 in TCMSP database, 51 in Herb database, 32 in HIT database, 18 in PubChem database, 79 in PharmMapper database, 57 in SwissTargetPrediction database and 7 in STITCH database. A total of 196 potential TanIIA-targeted genes were collected after deduplication. As shown in Figure 2B , the TanIIA-TanIIA target network shows 197 nodes (196 targets and TanIIA), which represent the interaction between TanIIA and its potential therapeutic targets.

The number of related genes screened from the 4 databases were 609 in OncoDBHCC database, 6994 in GeneCards database, 245 in OMIM database and 48 in Drugbank database. After removing duplicates, 7169 genes were found.

The three sets of data obtained above, including TanIIA-targeted genes, HCC-related genes and differential genes, were screened out to have a total of 64 genes, and the Venn diagram was drawn by R package ggplot2 ( Figure 2C ). These 64 genes were defined as TanIIA-related genes. The TanIIA-related genes were introduced into String to construct a PPI network, and the free nodes were removed, resulting in a total of 57 targets ( Figure 2D ). The degree value was sorted by Cytoscape 3.8, and 40 key target proteins with degree value > 5 were screened out. The bar chart of corresponding degree values of target proteins is shown in Figure 2E . The top eight target proteins were ALB, JUN, MYC, SRC, ESR1, MMP9, PTGS2 and FOS.

KEGG/GO enrichment was performed on the 40 key target genes obtained above. On the premise of p < 0.05, the top 15 pathways and functional modules are shown in Figure 3 . From the results of KEGG ( Figure 3A, E ), it is found that the key targets of TanIIA in the treatment of HCC are mainly involved in the MAPK signal pathway, Transcriptional misregulation in cancer, Hepatitis B, TNF signal pathway, endocrine resistance and other signal pathways, indicating that the therapeutic effect of TanIIA on HCC may be realized through the above pathways. GO enrichment results suggested that the biological processes were mainly involved in steroid metabolic process, regulation of inflammatory response, response to oxidative stress, ERK1 and ERK2 cascade and response to molecule of bacterial origin, etc. ( Figure 3B, F ). The cell components were associated with nuclear chromatin, membrane region, membrane microdomain, membrane raft and protein-DNA complex, etc. ( Figure 3C, G ). The molecular functions were related to heme binding, tetrapyrrole binding, oxidoreductase activity, monooxygenase activity, DNA-binding transcription activator activity, etc. ( Figure 3D, H ). Taken together, these indicate that TanIIA mainly affects cell proliferation, apoptosis and migration through its effects on oxidative stress, hormone metabolism, inflammatory immunity, and so on, so as to have therapeutic effects on HCC.

To further explore the prognostic value of TanIIA-related genes in HCC, KM analysis and COX in TCGA-LIHC cohort were utilized to assess the prognosis significance of TanIIA-related genes. KM analysis results showed that 5 genes (ALB, JUN, SRC, ESR1 and MMP9) met the criteria of P < 0.05, indicating their associations with OS prognosis ( Figure 4A–E ). And among those genes, which were identified to be associated with OS, 3 genes (JUN, SRC and MMP9) were considered as risk factors with HRs >1, while another 2 genes (ALB and ESR1) were protective factors with HR <1 ( Figure 4F ). In addition, LASSO Cox regression analysis was performed to construct a prognosis prediction model. In accordance with the optimum λ value with ten-fold cross-validation, the 5 TanIIA-related gene signatures were selected, and a 5-gene prediction model was established ( Figures 4G, H ).

TCGA-LIHC cohort was used as the training set. Moreover, it was also employed as testing set for internal cross validation. The risk score was derived from the following formula: riskscore = (-0.0324)×ALB expression +(0.0519)×JUN expression +(0.0524)×SRC expression +(-0.1366)×ESR1 expression +(0.0588)×MMP9 expression. Based on the risk scores, HCC patients in TCGA database were stratified into low- and high-risk subgroups. The area under the time-dependent ROC curve (AUC) were 0.680 for 1-year, 0.627 for 3-years, and 0.632 for 5-years ( Figure 4I ). And the KM curve revealed that the low-risk group had better prognostic than the high-risk group. ( Figure 4J ). Besides, it was also analyzed for distribution of selected gene expressions and risk scores in the TCGA cohort ( Figure 4K ).

The expression of the five prognostic TanIIA-related genes were further analyzed. The analysis results demonstrated that the genes were differentially expressed in HCC and normal samples ( Figures A–E ). And the correlation among the 5 gene expression was shown in Figure 5F .

Then the relationship between drug sensitivities of Sorafenib and expression levels were evaluated. Among five genes, ALB, ESR1 and SRC was more likely to serve as potential targets for sorafenib. ALB (r=0.273, p>0.01, Figure 6A ) and ESR1 (r=0.220, p>0.01, Figure 6B ) was negatively correlated with sorafenib IC50 value. Meanwhile, SRC was negatively (r=-0.359, p>0.01, Figure 6C ). The results were in accordance with the results of Spearman correlation analysis of the 5 prognostic TanIIA-related genes ( Figure 5F ). Therefore, we chose ALB, ESR1 and SRC as our core genes for further investigation.

The statistical correlation between immunity and expression levels of prognostic TanIIA-related genes were also clarified, including immune cell infiltration and immune scores. It was shown that Neutrophils cells, DC cells, Treg cells, Th17 cells, Tgd cells and Cytotoxic cells, etc. were significantly positively correlated with the expression of ALB and ESR1 ( Figures 6D, E ). While, most of the above positive-correlated immune cells, such as DC cells, Treg cells, Th17 cells and Tgd cells shown a negative correlation with SRC ( Figure 6F ). Moreover, the expression of ALB and ESR1 and the StromalScores and ESTIMATEScores were positively correlated ( Figures 6G, H ), while SRC expression shown no statistically correlation ( Figure 6I ). These results corroborated that the core genes are correlated to the development and progression of HCC.

Sybyl-X was utilized to locate the binding sites and evaluated the binding score values of the compounds. According to the molecular docking results, TanIIA had well protein affinities with core genes relevant proteins ( Figure 7 ). And docking scores were>4, suggesting relatively stable binding activities as well.

Three prognostic model genes were predicted by network pharmacology and machine learning experiments. In order to verify the effectiveness and reliability of these genes in the treatment of HCC, a cell model experiment was conducted to verify the results. Hep3B and HepG2 cells were intervened with different concentrations of TanIIA for 24 h, and the cell survival rate was examined by CCK-8 assay. It turned out that with the increase of TanIIA concentration, the cell viabilities of both cell lines were more inhibited, which indicated the correlation with the dose ( Figure 8A ). After analyzing the experimental data, the IC₅₀ values of TanIIA for Hep3B and hepG2 were respectively calculated to be 42.45 μmol and 36.71 μmol. Therefore, the three different concentrations (low, medium and high dosing) of TanIIA were determined for subsequent experiments, whose administration concentrations were 20 μmol, 40 μmol and 80 μmol, respectively.

Subsequently, whether TanIIA influenced HCC cell apoptosis or not was determined via FCM. The results of apoptosis are shown in Figures 8D, E . The total apoptosis rate = early apoptosis value + late apoptosis value. With the increase of TanIIA concentration, the apoptosis rate gradually increased in Hep3B cells ( Figure 8B ) as well, from (5.49 ± 0.91) % in the control group to (8.4 ± 5.57) %, (16.63 ± 2.24) % and (27.72 ± 4.22) %. And in HepG2 cells ( Figure 8C ), the apoptosis rates increased from (3.84 ± 2.89) % to (12.86 ± 1.93) %, (21.69 ± 9.11) % and (25.72 ± 4.31) %, respectively. Compared with the control group, the induction effect of TanIIA on Hep3B cells at the concentration of 80 μmol was significantly different (p<0.01), and there was a significant difference in the induction of HepG2 cells at the concentration of 40 and 80 μmol (p<0.01). Interestingly, necrotic cells were also gradually increased, especially in HepG2 cells. The number of necrotic cells was more than 20% at 80 μmol. This indicated that TanIIA may induce apoptosis and necrosis in HepG2 cells at the same time.

As demonstrated in Figure 8F-I , after intervened with TanIIA, the number of invading cells was suppressed in a dose-dependent manner. Hep3B cells intervened with 80 μmol TanIIA showed remarkably inhibition in invasion ( Figure 8H , p<0.01). Similarly, it found that the invasion ability of HepG2 cells was weakened after treatment with 80 μmol TanIIA ( Figure 8I , p<0.05). These results confirmed that Tan IIA can inhibit cell invasion of HCC.

Based on the above results of FCM and transwell assay, the effect of TanIIA on HCC cells was further investigated. Therefore, the expression level changes of apoptotic proteins (Bax, Bcl-2) and invasion protein (MMP9) were detected. As shown in the Figure 9A , the expression of Bax showed an increase in both Hep3B and HepG2 cells intervened with TanIIA for 24h. Whilst, the expression levels of Bcl-2 and MMP9 both decreased as well. Compared with the control group, the expression levels of TanIIA-intervened group were changed with statistically significant difference ( Figure 9B , p<0.05).

Taken together, these results prove that TanIIA contributes to inhibiting cell proliferation, inducing apoptosis and preventing cell invasion in HCC. It might play an important role in the anticancer progression of HCC.

Next, western blot analysis was subjected to detect the three core genes. Results indicated that the two genes, ALB and ESR1, were evoked by TanIIA ( Figures 9C, D ). And the expression levels of SRC and p-ERK1/2 were down-regulated ( Figures 9C, D ). These statistically significant results suggest that TanIIA simultaneously influences the prognostic TanIIA-related genes in HCC cells.