Esculetin (CAS:531-75-9) was purchased from National Institutes for Food and Drug Control (China) (purity ≥ 99.18%). PGK2 (ab123166), GPD2(ab153790), GPI (ab208311), PGK2 monoclonal antibody (ab183031), and GPD2 monoclonal antibody (ab182144) were purchased from Abcam (Cambridge, UK). GPI monoclonal antibody (8866S), Casepase-3 antibody (9662), BAX antibody (2772T), and Bcl-2 Rabbit mAb (3498T) was obtained from Cell Signaling Technology (Boston, USA). Bovine serum albumin (BSA) was purchased from Sigma (St. Louis, MO, USA). TruSeq™ RNA sample preparation Kit from Illumina (San Diego, CA). SuperScript double-stranded cDNA synthesis kit (Invitrogen, CA).

The HepG2 cells were inoculated into six-well plates at a concentration of 5 x 10⁵/ml. There are two groups: HepG2 cells without esculetin and HepG2 cells containing esculetin (n=3). According to the relevant literature (Wang et al., 2015), esculetin was added at 50 µM to each well and incubated for 24 h at 37°C with 5% CO₂.

According to the manufacturer’s instructions (Invitrogen), getting the total RNA from HepG2 cells. RNA quality was determined. Sequencing libraries were constructed using only high-quality RNA samples (OD260/280 = 1.8~2.2).

According to the TruSeq™ RNA sample preparation Kit of Illumina, we prepared the RNA-seq transcriptome library by using 5 μg of total RNA. For a short period of time, we isolated mRNA by oligo(dT) beads and RNA was fragmented through a fragment buffer. The SuperScript double-stranded cDNA synthesis kit was used to synthesize double-stranded cDNA. The 200–300 bp cDNA target fragments size on 2% low range ultra agarose were selected, and 15 PCR cycles were amplified using Phusion DNA polymerase (NEB). Illumina HiSeq X Ten (2 × 150 bp read length) was sequenced using the paired-end RNA-seq sequencing library after quantification by TBS380.

The expression level of each transcript was calculated by the number of segments per kilobase of exon per million mapped reads (FRKM) method, in order to identify differential expression genes (DEGs) between two different samples. RSEM (http://deweylab.biostat.wisc.edu/rsem/) (Li and Dewey, 2011) was used for quantitative genetic abundances. “The R statistical software package EdgeR (Empirical analysis of Digital Gene Expression in R, http://www.bioconductor.org/packages/2.12/bioc/html/edgeR.html)” (Robinson et al., 2010) was used to assess differences. Functional enrichment was performed using gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG). “GO functional enrichment and KEGG pathway analysis were performed by Goatools (https://github.com/tanghaibao/Goatools) and KOBAS (http://kobas.cbi.pku.edu.cn/home.do)” (Xie et al., 2011).

PharmMapper is a kind of free use of network tools, the pharmacophore mapping method is used to identify potential targets from the results of transcriptome. Using MalaCards and GeneCards for further screening. Uploaded all potential targets to the DAVID database. Using Gene Ontology and KEGG pathway to study the functions and pathways of enrichment.

The network was built by Cytoscape 3.6.1 software. Inputted the active compound and target gene as the node, and if there was a connection between the two nodes, added an edge to show the connection.

The “network analysis” function was used to analyze the network. Based on the KEGG analysis, key targets of the selected path with the largest counts were analyzed. According to the literature and database, we obtained the proteins suitable for target analysis (Hopkins, 2007; Shao and Zhang, 2013).

Just like above, the HepG2 cells were treated or not treated with 0–200 µM esculetin (0.9% NaCl solution containing 1% EtOH) for 1 h, and the medium was collected. The concentrations of glucose and lactate in the culture medium were determined by using the Model680 automatic enzyme labeling instrument from BIO-RAD Company, USA. Glucose consumption and lactate production were calculated according to the difference between the concentrations at the start of the culture and the appropriate time. At the same time, in order to exclude the influence of cell viability on the experiment, we tested the cell viability under different incubation times and different dose concentrations (Chen et al., 2012).

Esculetin was analyzed with the proteins (PGK2, GPD2, GPI, and the BSA for negative control) labeled with the Labeling Kit. LED power was set to 20%.

Monolith Nano Temper (NT) 115 was used for measuring molecular interactions. Based on the fluorescence intensities that were determined through MST measurements, and then Kd values calculated (van den Bogaart et al., 2012; Batoulis et al., 2016; Vinothkannan Ravichandran et al., 2018).

The interaction between GPD2 and GPI with higher degree in network were detect by MST experiments. GPD2 (20 nM) was used with a gradient of 16 concentrations of unlabeled GPI. As before, cell lysate were collected to incubated overnight with anti-GPD2 antibody or normal mouse IgG (as a control) and protein A/G beads at 4°C. The co-immunoprecipitated proteins were detected by mass spectrometry and western blot. Then, bioinformatics analysis was carried out for the different genes between the two groups.

Male BALB/c mice (18–22 g) were obtained from the Hubei Provincial Center for Disease Control and Prevention (Wuhan, China). All procedures have been approved by the Animal Care and Use Committee of Institute of Materia Medica, People’s Republic of China and the H22 cells commonly used in liver-cancer modeling in normal mice were selected. H22 cells (1 x 10⁷) were injected subcutaneously (s.c.) into the right flanks of the mice, according to the literature (Zhang et al., 2019). In our preliminary experiments, we found that the toxicity of esculetin in vivo was negligible at 50 mg/kg, so mice were divided into three groups of eight mice, and intraperitoneal treatment was performed as follows: control group (daily received physiological saline) and esculetin-treatment group (daily received 25 and 50 mg/kg esculetin for 15 d). Antitumor activity was evaluated by weight. Also, PGK2, GPD2, and GPI expression was measured using western blot. Different protein samples were loaded onto SDS-PAGE and then electrotransferred to PVDF membranes. The membrane was then sealed with skim milk and incubated with antibodies. Finally, the results were then visualized by the ECL detection system.

The DEGs between two samples are represented in a volcanic map (Figure 1A). These esculetin-induced DEGs were classified into different biological process by the KEGG analysis. The top 20 clusters include the following (Figure 1B): glycolysis/gluconeogenesis, fatty acid degradation, glycerolipid metabolism, and fructose and mannose metabolism. The GO analysis showed that the DEGs were significantly enriched in glucose 6-phosphate metabolic processes, acetyl-CoA biosynthesis, and glycosyl compound metabolic processes (Figure 1C). The enriched processes of DEGs focused on the processes related to glycolysis that are specially utilized by cancer, which may be the mechanism of anticancer effect of esculetin. We further verified this in the following experiments, for more explanation of the figure, see the legend (Table S1).

KEGG analysis showed that 42 of the genes were enriched (97.7%), including 38 pathways, the target genes were significantly correlated with 33 of them (P ≤ 0.05) (Figure 2A). These pathways have the highest target genes involvement: Metabolic pathways (35, 81.4%), glycolysis/gluconeogenesis (33, 76.6%), carbon metabolism (24, 55.8%), pyruvate metabolism (12, 27.9%), and fructose and mannose metabolism (10, 23.3%). GO enrichment analysis showed that these genes involved in the CC (Figure 2C), MF (Figure 2D), and BP (Figure 2B) targets were 44 (100%). CC enrichment indicated that targets involved in those areas: cytosol (27, 62.8%), extracellular exosome (24, 55.8%), and mitochondrion (13, 30.2%). MF enrichment mainly involved target genes in the interactions: ATP binding (12, 27.9%), identical protein binding (6, 14%), and aldehyde dehydrogenase (NAD) activity (5, 11.6%). BP enrichment mostly involved targets in these processes: glycolytic process (18, 41.9%), canonical glycolysis (16, 37.2%), and gluconeogenesis (13, 30.2%).

The target genes of the top 10 pathways and the compound were used to construct the network (Figure 3). The network diagram showed a synergistic effect on multiple targets for the anti-cancer effect of the esculetin. When esculetin exhibits an anti-cancer effect, GPI, GPD2, and PGK2 with a higher degree were selected for the following experiments.

Inhibitions of PGK2, GPD2, and GPI, were reflected by changes in glucose consumption and lactate production in the cell. Cellular glucose consumption, lactate production, and cell death were determined, after cells were treated with 0–200 µM esculetin within 1 h. The results showed that esculetin inhibited glucose consumption and lactate production in cells (Figures 4B, C). Considering that esculetin may kill cells and might lead to a decrease in glycolytic flux, we studied cell mortality, which were negligible for 1-h incubation (Figure 4A). Consistent with its inhibition on glycolytic flux, HepG2 cells were toxic under given experimental condition (Figures 4D, E).

The interactions between inhibitor and proteins were assayed by the MST. Multiply all values produce a relative change per thousand fluorescence. The Kd value of PGK2, GPD2 and GPI are shown as Figure 5 and the Kd of negative control BSA is failed.

Using network pharmacology, we screened GPD2 and GPI with the high degree of affinity. We further established the binding between protein combinations and calculated the Kd by MST. According to the association curve, proteins have a conformational change with a S-shaped (Figures 6A, B). The result of Co-IP also indicated that the proteins bind well to each other (Figure 6C) and the differentially expressed genes were concentrated in glycolysis pathways (Figures 6D, E) (for proteomics data, Table S2).

The antitumor activity of esculetin was evaluated by a xenograft of H22 cells in BALB/c mice. The tumor weight increased significantly in the model group compared with that in the esculetin treatment group (Figure 7A). Target protein measurement of the tumor tissue revealed an increase in PGK2, GPD2, and GPI in response to the esculetin treatment (Figures 7B–D). This may be a compensatory response when the proteins binding to small molecules were inhibited and thus the glycolysis-related process is blocked. The up-regulation of caspase-3, Bax, and the down-regulation of Bcl-2 indicated that the drug induced the apoptosis of tumor cells.