The Bel-7402 and HepG2 cell lines, both widely used for hepatocellular carcinoma research, were acquired from the American Type Culture Collection (ATCC) and cultured under sterile conditions to ensure cell viability and purity. The cells were maintained in a culture flask with Dulbecco’s Modified Eagle’s Medium (DMEM, Gibco, Cat. No. 11-965-092, USA), supplemented with 10% fetal bovine serum (FBS, Gibco, Cat. No.11560636, USA) and 1% penicillin-streptomycin (ThermoFicher Scientific, Cat. No.15323671, USA). The cultures were maintained in a humidified incubator at 37°C with 5% CO2. Passaging of the cells was carried out when the cells reached approximately 80-90% confluence, 0.25% Trypsin-EDTA (ThermoFisher Scientific, Cat. No. 25200072) was used to digest cells, the cell suspension was then transferred to new culture flasks for further experiments.

To assess the cytotoxic effects of Metformin and Ferrostatin-1 on Bel-7402 and HepG2 cells, MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay was employed to determine the cell viability and proliferation. Briefly, a total of 10000 cells in 200 μL medium were pipetted into 96-well plates to grow for 24h. They were then pre-treated with 1 µM Ferrostatin-1 for 24 hours (21) before exposure to Metformin (Santa-Cruz, Cat. No. 1115-70-4) (0, 1, 2, 5, 10 and 20mM, diluted in phosphate buffer saline) for 24, 48, or 72 hours. After treatment, 20 μL of MTT reagent was added to each well, and the cells were incubated for 4 hours at 37°C. Then the medium was discarded, and formazan precipitate was dissolved in 150 μL DMSO. Absorbance at 570 nm was measured by a plate reader (Tecan Spark^®, Männedorf, Switzerland). Each experimental group was conducted in triplicate.

Total cellular proteins were extracted using RIPA buffer (ThermoFisher Scientific, Cat. No. 89900), supplemented with a cocktail of protease and phosphatase inhibitors (Roche) to prevent degradation and dephosphorylation of target proteins. Extracts from each sample were reduced with 2mM DTT for 1 hour at 56°C, and subsequently alkylated with sufficient iodoacetamide for 1 hour at room temperature in the dark. Then 4 times the volume of precooled acetone was mixed with samples by well vortexing and incubated at -20°C for at least 2 hours. Samples were then centrifuged and the precipitation was collected. After washing twice with cold acetone, the pellet was dissolved by dissolution buffer, which contained 0.1 M triethylammonium bicarbonate (TEAB, pH 8.5) and 6 M urea. Protein concentration was determined again by Bradford protein assay.

Digestion of 5 mg of protein was performed with Trypsin Gold at a 1:50 enzyme-to-substrate ratio for 16 hours at 37°C. Peptides were desalted using a C18 cartridge and dried by vacuum centrifugation. For phosphopeptide enrichment, peptides were redissolved in 250 mM acetic acid with 30% acetonitrile, adjusted to pH 2.5-3.0, and processed using phos-select iron affinity gel according to the manufacturer’s protocol. The bounded peptides were eluted, dried and desalted using peptide desalting spin columns (Thermo Fisher, 89852).

Peptides were analyzed using an EASY-nLC™ 1200 UHPLC system (Thermo Fisher) coupled with an Orbitrap Q Exactive HF-X mass spectrometer (Thermo Fisher) operated in data-dependent acquisition (DDA) mode and positive polarity. Full MS scans were acquired at a resolution of 60,000 (at m/z 200) over a scan range of 350–1,500 m/z. The automatic gain control (AGC) target for MS1 was set to 3 × 10⁶, with a maximum injection time of 50 ms. The top 20 most intense precursor ions were selected for higher-energy collisional dissociation (HCD) with a normalized collision energy (NCE) of 27%. MS/MS scans were acquired at a resolution of 15,000 (at m/z 200), using an isolation window of 1.6 m/z, an AGC target of 1 × 10⁵, and a maximum injection time of 45 ms. Dynamic exclusion was enabled with a duration of 30 seconds to avoid repeated selection of precursor ions.

The resulting spectra were searched against a protein database using Proteome Discoverer 2.2 (PD 2.2, Thermo), with carbamidomethylation as a fixed modification, and oxidation of methionine, phosphorylation of serine (S), threonine (T), and tyrosine (Y), and N-terminal acetylation as variable modifications. Protein identification was performed at a false discovery rate (FDR) < 1.0%, and label-free quantification was applied. The protein quantitation results were statistically analyzed by the Mann–Whitney test. Differentially expressed proteins (DEPs) were defined based on fold-change (ratio > 4 or < 0.25) and statistical significance (P < 0.05).

Functional analysis of proteins and DEPs was conducted using Gene Ontology (GO), InterPro (IPR), COG (Clusters of Orthologous Groups), and KEGG (Kyoto Encyclopedia of Genes and Genomes) databases. GO and IPR analysis was conducted using the InterProScan-5 program against the non-redundant protein database (including Pfam, PRINTS, ProDom, SMART, ProSiteProfiles, and PANTHER). STRING-db (http://string.embl.de/) was used to predict potential protein–protein interactions based on related species. Motif enrichment analysis was conducted using the motif-x algorithm, focusing on a 7-amino acid window surrounding each phosphosite (occurrence > 20, P < 10⁻⁶), and visualized with WebLogo. Phosphorylation site–kinase relationships were predicted using the NetPhorest algorithm.

Total cellular proteins were extracted using RIPA buffer (ThermoFisher Scientific, Cat. No. 89900), supplemented with a cocktail of protease and phosphatase inhibitors (Roche, Cat. No. 04693132001) to prevent degradation and dephosphorylation of target proteins. Protein concentrations were quantified using the BCA Protein Assay Kit (ThermoFisher Scientific, Cat. No. A55860), ensuring accurate loading of samples. 50 μg of each protein extract was resolved by SDS-PAGE and electrophoretically transferred onto a nitrocellulose membrane. The membranes were then incubated with primary antibodies specific for ATG9 (Abcam, ab108338, diluted 1:1000), ATG5 (Abcam, ab108327, diluted 1:1000), ATG16 (Abcam, ab188642, diluted 1:1000), P62 (Abcam, ab109012, diluted 1:10000), Beclin-1 (Abcam, ab207612, diluted 1:2000), Ic3b (Abcam, ab231078, diluted 1:1000), SLC7A (Abcam, ab236669, diluted 1:500), GPX4 (Abcam, ab125066, diluted 1:1000), ASCL4 (Abcam, ab155282, diluted 1:10000), and β-actin (Abcam, ab8227, diluted 1:1000) overnight at 4°C to ensure efficient antigen-antibody binding. After thorough washing with Tris-buffered saline with Tween (TBST) to remove unbound primary antibodies, the membranes were incubated with an HRP-conjugated secondary antibody (Bio-Rad, diluted 1:2000) for 1 hour at room temperature to amplify the signal. To visualize the immunoreactive bands, the membranes were developed using a chemiluminescent substrate (ThermoFisher Scientific, Cat. No. 34080). The luminescent signal was detected by ChemiDoc™ Imaging System (Bio-Rad) and relative expression levels normalized to β-actin were calculated

The cell viability of two cell lines including Bel-7402 and HepG2 treated with Metformin of multiple concentrations was detected by MTT assay. The results showed that the cell viability of Bel-7402 cells indicated a negative correlation with Metformin concentration at different time points of 24, 48 and 72 hours. Among these, the Bel-7402 cells treated with 20mM Metformin at 72 hours exhibited the lowest cell viability compared with other groups, thus this concentration was selected for further experiment. Similarly, the results in HepG2 cells also represented the higher concentration of Metformin or longer duration of treatment, the lower cell viability of cells ( Figure 1 ). These results indicated the inhibitory effect of Metformin on the proliferation of Bel-7402 and HepG2.

To further explore the molecular regulation mechanism of Metformin in hepatocellular carcinoma cells, phosphoproteomics and bioinformatics analysis were performed. As shown in Figure 2 , the heatmap and volcano plot showed a distinct difference between the control group (Bel_0/Hep_0) and Metformin-treated groups (Bel_20/Hep_20) either in Bel-7402 or HepG2 cells. Furthermore, subcellular localization analysis revealed that the alternation of DEPs was mainly located in the nucleus, with 44.58% in Bel-7402 and 47.95% in HepG2 cells, followed by cytoplasm proteins.

The function of DEPs between control and Metformin-treated groups in two cell groups was further analyzed by Gene Ontology analysis. Specifically, the enriched GO analysis showed that the top 5 terms in Bel-7402 cells were pyrimidine nucleotide biosynthetic process, chemical homeostasis, GTP biosynthetic process, UTP biosynthetic process and CTP biosynthetic process in Biological Process (BP); membrane region, endoplasmic reticulum, cytoplasmic part, membrane raft, whole membrane in Cellular Component (CC); and ion channel activity, protein phosphatase regulator activity, calcium ion transmembrane transporter activity, nucleoside diphosphate kinase activity, phosphatase binding in Molecular Function (MF) ( Figure 3A ). The KEGG pathway analysis of Bel-7402 cells suggested the significant difference in the pathway between Bel_0 and Bel_20 was mainly focused on the biosynthesis and metabolism regulation, such as Ubiquinone and other terpenoid-quinone biosynthesis, Glycine, serine and threonine metabolism, and Insulin secretion, etc ( Figure 3B ).

As for HepG2 cells, the top 5 enriched terms analyzed by GO were DNA metabolic process, organic substance transport, DNA replication, macromolecule localization and biosynthetic process in BP; plasma membrane protein complex, protein serine/threonine phosphatase complex, mitochondrial membrane part, clathrin coat of trans-Golgi network vesicle, clathrin coat of coated pit in CC; and ATPase activity, protein domain specific binding, phosphatidylinositol phospholipase C activity, phosphoric diester hydrolase activity and microtubule motor activity in MF ( Figure 3C ). Moreover, the KEGG pathway analysis of HepG2 cells suggested Pentose and glucuronate interconversions, Citrate cycle (TCA cycle), Arginine biosynthesis, Pyruvate metabolism and mismatch repair, suggested a primary alternation of biosynthesis and metabolism regulation in HepG2_0 and HepG2_20, similarly to the GO data in Bel-7402 cells ( Figure 3D ).

Notably, KEGG analysis revealed that differentially expressed protein enriched in ferroptosis pathway, as well. Furthermore, several studies shown that ferroptosis is closely interconnected with various molecular pathways and cellular processes, including immunological signaling, transcription factors, other cell death modalities, and metabolic pathways, which were all mentioned in our phosphoproteomics results above ( Figure 3 ). This complex network of interactions highlights that ferroptosis is a key molecular process involved in the Metformin inhibitory effect on hepatocellular carcinoma (22–24).

As represented in Figure 4 , the downregulation of SLC3A2, and the upregulations of SLC7A11 and GCL signals were found to participate in the ferroptosis process, these proteins were also found to be a significant difference between Bel_0 and Bel_20 groups. Consistently, several DEPs related to ferroptosis were also determined in HepG2_0 and HepG2_20, such as the overexpression of SLC3A2, SLC7A11, TF, and downregulation of PCBP2 and HO-1.

MTT assay demonstrated that metformin could significantly suppress the cell viability of Bel-7402, while its inhibitory role could be partly reversed by ferrostatin-1 (all p<0.05) ( Figure 5A ), also known as Fer-1, a potent and selective inhibitor of ferroptosis. In line with Bel-7402 cells, there is a dramatic difference between Fer-1 and Metformin treatment groups in HepG2 cells (p<0.05), and the suppressive impact of Metformin could be partly eliminated by Fer-1 (p<0.05), as shown in Figure 5B .

Based on the partially reversed MTT results and the fact that ferroptosis closely interacts with autophagy (24), WB was performed to detect the ferroptosis and Autophagy-related protein expressions in HepG2 and Bel-7402 cells. The expression levels in HepG2 were shown in Figures 6A, C , the data showed that the Autophagy-related proteins including ATG9, ATG5, P62 were decreased in Metformin group, and could be recused by Fer-1, while ATG16 and Ic3b did not show an obvious rescue effect after Fer-1 intervention in Metformin-treated HepG2 cells. The ferroptosis proteins such as SLC7A and GPX4 showed a significant inhibition after Metformin treatment, and could be upregulated after Fer-1 intervention. No significant change was noticed in ASCL4 between Metformin and Metformin+Fer-1 groups.

As for Bel-7402 cell line, its expression results were shown in Figures 6B, D , the protein expressions of ATG5, P62, Ic3b were downregulated in Metformin group, and were rescued by Fer-1 treatment. However, ATG9 and ATG16 showed opposite results that the Metformin could elevate their expression whereas this effect could be partly reversed by Fer-1 intervention. Intriguingly, ferroptosis-relevant proteins including SLC7A, GPX4 and ASCL4 showed different tendencies in Bel-7402 cells, indicating an intrinsic interaction of the ferroptosis process.