Reagents and antibodies are as follows: sorafenib (MedChemExpress, Cat# HY-10201); lenvatinib (MedChemExpress, Cat# HY-10981); fetal bovine serum (Gemini Bio, Sacramento, CA); Lipofectamine 3,000 (Invitrogen); Dulbecco’s Modified Eagle Medium (DMEM) (Gibco, Grand Island, NY); Fluoromount with 4′, 6- diamidino-2-phenylindole (DAPI) (Sigma-Aldrich, St. Louis, MO); Puromycin (Solarbio Life Science, Beijing, China); GAPDH (Proteintech, Cat# 60004-1-Ig); MISP (Proteintech, Cat# 26338-1-AP); CHMP2B (Proteintech, Cat# 12527-1-AP); TMSB4X (Proteintech, Cat# 19850-1-AP); IL-18 (ABclonal, Cat# A1115); EFEMP1 (Abcam, Cat# ab256457).

Retrospective collection of tissue paraffin embedded samples from HCC patients admitted to Fujian Provincial Hospital from January 2023 to December 2023. This research protocol has been approved by the Ethics Committee of Fujian Provincial Hospital (K2023-05-016).

The Huh7 cell lines (RRID: CVCL_0336, JCRB0403, Japan) were cultured in DMEM, supplemented with 10% FBS and 100 U/mL Penicillin/Streptomycin, and maintained at 37°C in a 5% CO₂ atmosphere. The cells underwent routine testing for mycoplasma contamination, which confirmed their freedom from contamination.

The Huh7/SR and Huh7/RR cell lines, which are resistant to sorafenib or lenvatinib, were created in a previous study (Wang et al., 2023b; Leung et al., 2023). Briefly, Huh7 cells were cultured with gradually increasing doses of sorafenib or lenvatinib. Both Huh7/SR and Huh7/RR cells were cultured at a constant concentration of 10 μM.

si MISP (target sequence: TTC​CGT​TTC​TAT​CTT​CCT​TTA​GA), si CHMP2B (target sequence: AAG​AAA​ACC​GTG​GAT​GGA​ATT​AG), si IL-18 (target sequence: AAC​TAT​TTG​TCG​CAG​GAA​TAA​AG), si TMSB4X (target sequence: TAG​CTG​TTT​AAC​TTT​GTA​AGA​TG), si EFEMP1 (target sequence: CGC​ACA​GAT​TCA​CAA​TGT​TGA​AA) and scrambled control si RNA (si Control), were purchased from Tsingke Biotechnology Co., Ltd. All siRNA transfections were performed using Lipofectamine 3,000 according to the manufacturer’s protocol.

The CCK-8 reagent was used to test cell viability in accordance with the manufacturer’s instructions. In brief, the 96 well plate had 5,000 cells per well. Next, 10 μL of CCK-8 reagent was added to each well. The absorbance of every well was measured at 450 nm with a microplate reader (BioTek, Winooski, VT).

Immunoblotting was based on our previous research (Wang et al., 2023a). The total protein samples (25 μg) were separated through SDS-PAGE were separated via SDS-PAGE and transferred onto PVDF membranes (Biosharp, Hefei, China). Following this, the membranes were blocked with 5% bovine serum albumin (BSA) at room temperature for 1 h. The primary antibodies were applied to the membranes and incubated overnight at 4°C. After this step, secondary antibodies were added and incubated at room temperature for 1 h. The resulting bands were detected and visualized using a Hypersensitive ECL Chemiluminescence Kit (NcmECL Ultra, ABP Biosciences, Beltsville, MD, USA). The study utilized several primary antibodies: anti-MISP (1:1,000), anti-CHMP2B (1:1,000), anti-IL-18 (1:1,000), anti-TMSB4X (1:1,000), and anti-EFEMP1 (1:1,000), anti-GAPDH (1:5,000).

The EdU Cell Proliferation Kit (Solarbio, China) was utilized to quantify cell proliferation as per the guidelines furnished by the manufacturer. Upon combining with a fluorescent azide, proliferating cells emitted a vibrant red fluorescence that was visualized under a fluorescent microscope.

4D-DIA quantitative proteomics was detected and analyzed by Novogen Co., Ltd. The software used for integrating metabolomics and proteomics data were presented in Supplementary Table S1. The steps are as follows.

Untargeted metabolomics was detected and analyzed by Novogen Co., Ltd. The steps are as follows.

The results are expressed as mean ± SD and conducted using GraphPad Prism V.8. For normally distributed data with homogeneous variance, use unpaired t-tests to compare two samples. The comparison between multiple groups was conducted using one-way analysis of variance. A p-value <0.05 is considered statistically significant.

We produced Huh7 sorafenib resistant and lenvatinib resistant cell lines (Huh7/SR, Huh7/RR), respectively. Compared to the parental Huh7 cells, both Huh7/SR and Huh7/RR cells have higher IC₅₀ values (Figure 1A). We next used the EdU proliferation detection kit to compare the number of EdU positive cells between parental cells and drug-resistant cells. It was found that compared to parental cells, both sorafenib and lenvatinib resistant cells had fewer Edu positive cells (Figures 1B, C). This indicates a decrease in the proliferation ability of drug-resistant cells. To evaluate the co-resistance mechanism of these two types of drug-resistant cells, we used 4D-DIA quantitative proteomics technology to explore the differences in protein expression. Both Huh7 and Huh7/RR, as well as Huh7 and Huh7/SR, have correlation coefficients greater than or equal to 0.94 (Figure 1D).

Cluster heatmaps display differences in protein expression patterns between two drug-resistant cells and parental cells (Figure 2A). In comparison to Huh7 cells, Huh7/SR cells displayed an increase in 972 protein expressions and a decrease in 1,051 protein expressions. Similarly, Huh7/RR cells showed an increase in 1,071 protein expressions and a decrease in 1,072 protein expressions (Figures 2B, C). Compared to Huh7/SR cells, Huh7/RR cells upregulated 504 proteins and downregulated 389 proteins (Supplementary Figure S1). Notably, we identified the top 10 upregulated and downregulated proteins. Co-upregulated proteins included MISP, CHMP2B, IL-18, TMSB4X and EFEMP1, while co-downregulated proteins comprised of IFITM3, CA4, AGR2 and SLC51B (Figures 2B, C). Subcellular localization analysis of proteins found that they are mainly concentrated in the nucleus, cytoplasm, mitochondrion and plasma membrane (Figures 2D, E). Venn diagram showed that there are 1,315 common differentially expressed proteins in both groups (Figure 2F). These data indicate that many proteins in drug-resistant cells have undergone changes, indicating that the protein profile has been reconstructed.

The gene ontology (GO) enrichment analysis revealed that, in comparison to Huh7 cells, both Huh7/SR and Huh7/RR cells exhibited shared differential enrichment signals. These signals were primarily associated with integral components of the plasma membrane, extracellular space, extracellular region, basolateral plasma membrane, apical plasma membrane, chaperonin-containing T-complex, calcium-dependent phospholipid binding, and immunological synapse (Figures 3A, B). KEGG pathway analysis showed that the common differential enrichment signals in Huh7/SR and Huh7/RR cells primarily involved steroid hormone biosynthesis, cell adhesion molecules, mucin type O-glycan biosynthesis, and glycosphingolipid biosynthesis—lacto and neolacto series (Figures 3C, D). Furthermore, the structural domain enrichment analysis revealed that the common differential structural domains in Huh7/SR and Huh7/RR cells were primarily associated with immunoglobulin-like folds and subtypes, immunoglobulin-like domains and their superfamily, immunoglobulin subtype 2, chaperone tailless complex polypeptide 1 (TCP-1), chaperonin TCP-1, conserved sites, groEL-like equatorial domain superfamily, and fibronectin type III (Figures 3E, F). Protein protein interaction analysis (PPI) showed that both Huh7/SR vs. Huh7 and Huh7/RR vs. Huh7 exhibited highly complex differentially expressed protein interactions, while the interaction of Huh7/RR vs. Huh7/SR was relatively reduced (Figure 3G). These results indicate that the Huh7/SR and Huh7/RR cell populations exhibit unique molecular characteristics, revealing rich pathways and structural domains involved in cellular signaling, biosynthesis, and immune responses, which can provide valuable insights for drug resistance in HCC treatment.

Because both proteins and metabolites are closely related to cellular function. We have constructed a proteomic profile of drug-resistant cells, and we next continue to construct a metabolomic profile of drug-resistant cells. PCA results showed differences in metabolomic separation trends among groups (Figure 4A). PC1 Scores suggested that the test samples were within the range of 3 standard deviations (SD) (Figure 4B). The clustering heatmap provided the differences in metabolites between two types of drug-resistant cells and parental cells (Figures 4C, D). The differential metabolite volcano plots displayed a visual representation of the statistical significance and magnitude of differences in metabolite abundance between groups. Compared to the Huh7 cells, the Huh7/SR cells had 176 metabolites elevated and 272 metabolites decreased (Figure 4E), while the Huh7/RR cell group had 89 metabolites elevated and 444 metabolites decreased (Figure 4F). These results can aid in understanding the metabolic pathways and regulatory mechanisms underlying cellular function and disease, providing a basis for potential biomarker discovery and therapeutic intervention.

To conduct in-depth analysis of differential metabolites, we conducted correlation analysis and enrichment analysis on differential metabolites. The correlation between differential metabolites indicated that amino acid and its metabolites were most significant among the three groups compared (Figures 5A–C). Metabolic Enrichment Analysis (MSEA) found that a large number of metabolic pathways were enriched, including: drug metabolism cytochrome P450, amino acid metabolism, fatty acid biosynthesis, pure metabolism, and glycolis/gluconeogenesis (Figures 5D, E). KEGG analysis of differential metabolites uncovered that the signal pathways jointly enriched in Huh7/SR and Huh7/RR cells involve the activation of chemical carcinogenesis receptors, the CGMP-PKG signaling pathway, and ABC transporters (Figures 5F, G).

By combining the proteomics and metabolomics analysis, it can be seen from the KEGG analysis results that the co-enriched signals of metabolism and proteomics in the two resistant cells are mainly reflected in: drug metabolism cytochrome P450, alanine, aspartate and glucose metabolism, ferroptosis, biosynthesis of amino acids, bill secretion, nucleotide metabolism, and pure metabolism (Figures 6A, B). The correlation analysis between proteins and metabolites revealed a intricate relationship within drug-resistant cells. Specifically, the third and seventh quadrants exhibited proteins and metabolites displaying a positive correlation, whereas the first and ninth quadrants indicated proteins and metabolites with discordant regulatory patterns (Figures 6C, D). Subsequently, we selected all proteins and metabolites that exhibited differential expression and utilized them to construct an O2PLS model. We then conducted a preliminary variable screening process where we identified variables that possessed a high correlation and weight in different data groups through load plots (Figures 6E, F). Through these analyses, we established a connection between differential metabolites and proteins, allowing for a comprehensive understanding of the underlying mechanisms of drug resistance.

To validate the protein family spectrum findings in drug-resistant cells, we selected HCC tissues from patients who underwent lenvatinib treatment and those with recurrence after such treatment. HE staining revealed necrosis in HCC tissue from patients treated with lenvatinib, along with an enhanced presence of immune infiltrating cells in para-carcinoma tissues and a notable increase in cancer cells in recurrent carcinoma tissue. These observations corroborate our understanding of the proteomic profiles in drug-resistant HCC cells (Figure 7A). The expression of MISP, CHMP2B, IL-18, TMSB4X, and EFEMP1 proteins in carcinoma tissue was higher than that in para-carcinoma tissues (Figure 7B). Survival analysis showed that high expression of MISP, CHMP2B, IL-18, TMSB4X, and EFEMP1 is not associated with poor prognosis in HCC (Figure 7C). Therefore, we speculated that these 5 proteins may not be related to tumor growth, but rather to drug resistance. We found that the expression levels of these 5 proteins in drug-resistant cells were significantly higher than those in parental cells, both in sorafenib resistant cells and lenvatinib resistant cells (Figure 7D). In addition, we used siRNA technology to knock down the expression of these 5 proteins one by one, and after being knocked down, the IC50 values of drug-resistant cells significantly decreased (Figures 7E, F). These results suggest that targeting these proteins will reduce drug resistance.