Huh7, SK-Hep1, Hep1–6 liver cancer cell lines were purchased from the American Type Culture Collection (ATCC) and authenticated by short tandem repeat (STR) profiling, which was performed by Qida Biotechnology (Shanghai, China). All liver cancer cell lines were cultured in Dulbecco’s Modified Eagle Medium (DMEM) (#10-013-CVRC, Corning, VA, United States) medium supplemented with 10% fetal bovine serum (FBS) (#10099-141, Gibco, NY, United States). All cells were maintained in a humidified incubator at 37°C with 5% CO₂.

HCC cells were seeded in low-adhesion 96-well plate at a density of 1,000 cells/well, with fresh culture medium replenished every three days. 10 days after culture, tumor spheres with diameter greater than 100μm were counted under a microscope. F12/DMEM supplemented with 1×B27, 20 ng/mL 20 ng/mL epidermal growth factor (EGF), 10 ng/mL fibroblast growth factor 10 (FGF10), and 10 ng/mL hepatocyte growth factor (HGF) and 1% PS was used as culture medium for sphere formation.

The ALDEFLUOR™ assay kit (#01700, STEMCELL Technologies, Vancouver, BC) was used for ALDH activity detection following the manufacturer’s protocol. A total of 2×10⁵ cells were centrifuged at 250g for 5 minutes, and the supernatant was discarded. The pellet was washed twice with the assay buffer. Cells was resuspended in 400μL of assay buffer mixed with 3μL activated ALDEFLUOR reagent, followed by dividing into 2 equal parts. 1 part were added with 3μL N, N-Diethylaminobenzaldehyde (DEAB) inhibitor to serve as the negative control, and the other part as the experimental one. Cells were incubated in dark at 37°C for 45 minutes. After incubation, cells were centrifuged at 250 g for 5 minutes, washed twice with the assay buffer, and resuspended in 300μL assay buffer for flow cytometry analysis within 4 hours on (#CytoFLEX S, Beckman Coulter Inc, CA, United States).

2×10⁵ cells were centrifuged at 250 g for 5 minutes, and the supernatant was discarded. Cells were washed twice with staining/washing buffer (1×Phosphate-Buffered Saline (PBS) with 2% FBS). Each sample was then resuspended in 300μL of buffer and incubated with 5μL of CD24 antibody (#PMG555428, Becton, Dickinson and Company, NJ, United States) on ice for 25 minutes. After incubation, cells were centrifuged at 250 g for 5 minutes, washed twice, resuspended and filtered for flow cytometry analysis within 4 hours on (#CytoFLEX S, Beckman Coulter Inc, CA, United States).

Tumor cells were lysed using RIPA lysis buffer (#P0013B, Beyotime, Shanghai, China), and protein concentration was quantified using the BCA Protein Assay Kit (#A55865, Thermo Fisher Scientific, MA, United States). The lysates were then subjected to SDS-PAGE and transferred onto PVDF membranes (#ISEQ00010, Millipore, Boston, United States). The membranes were blocked with 5% BSA for 1 hour at room temperature and incubated with primary antibodies at 4°C overnight. Subsequently, the blots were incubated with the horseradish peroxidase conjugated secondary antibody and developed by enhanced chemiluminescence.

Primary antibodies used were listed as following: Nanog (#4903), OCT4 (#2750), Sox2 (#2738), Sox9 (#D8G9H), GAPDH (#14C10), p-β-catenin-34/37 (#9561), β-catenin (#9562), p-GSK-3β (#9336), CyclinD1 (#2922), and TCF4 (#2569) were purchased from CST (United States), c-Myc (#9E10) was from Santa Cruz (United States).

Total RNA was extracted using TRI-Reagent (#TR118, Molecular Research Center, Inc, Cincinnati, OH, United States). Gene expression levels were measured using SYBR Premix Ex Taq II (Takara Bio, Shiga, Japan) on a 7,300 Real-Time PCR system (Applied Biosystems, Franklin Lakes, NJ, United States) with designed primers for target genes. The primers used in this study are listed in Supplementary Table S1 .

Cells in the logarithmic growth phase were counted, and 1,000 cells were seeded into each well of a 6-well plate. The cells were cultured for 10–14 days, with fresh medium changed every 3days. Colonies were fixed with 1mL 4% paraformaldehyde for 15 minutes, followed by staining with 0.1% crystal violet at room temperature for 20 minutes. Subsequently, stained colonies were washed with water until no residual dye remained. After the plate dried, images were taken. Finally, 10% acetic acid solution was added to dissolve the crystal violet for absorbance measuring at 530 nm using a microplate reader (#Multiskan SkyHigh, Thermo Fisher Scientific, Massachusetts, United States).

Cells were trypsinized, washed with PBS, resuspended in serum-free medium, and adjusted to a density of 2×10⁵/mL. 500 μL medium containing 20% FBS was added to the lower chamber, 100-200μL of cell suspension was added to the upper chamber, followed by incubation for desired time course. Afterward, the chambers were washed with PBS, and a cotton swab was used to remove non-migrated cells on the upper chamber side from the membrane. Migrated cells were fixed with 4% paraformaldehyde for 20 minutes, stained with 0.1% crystal violet for 30 minutes, and washed with PBS for three times. The chambers were dried at 37°C, and microscopic images were captured, with the number of migrated cells counted.

For the cell invasion assay, 20% Matrigel (#353097; Corning, NY, United States) diluted with serum-free medium was added to the upper chambers to mimic the extracellular matrix before the assay.

3.5×10⁴ cells were seeded onto coverslips in 12-well plates and cultured for 48–72 hours until an appropriate confluence was achieved. After washing with 1×PBS, the cells were fixed with 1mL 4% paraformaldehyde for 15 minutes. Next, the cells were permeabilized with 0.2% Triton X-100 for 10 minutes, followed by three washes with PBS. The cells were then blocked with 3% BSA for 15 minutes, followed by incubating with the primary antibody overnight at 4°C. Then, the cells were washed and incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG (H + L) (#ZF-0511, ZSGB-Bio, Beijing, China) for 1 hour in the dark, after washing, cells were mounted in DAPI-containing mounting medium (#P36941, Thermo Fisher Scientific, Massachusetts, United States) for imaging on a fluorescence microscope (FV3000, Olympus Corporation, Japan).

The IP assay was performed using an IP kit (P2197M, Beyotime, Shanghai, China). First, 300 μL of IP lysis buffer and 40 μL of Protein A magnetic beads were added to a sterile, enzyme-free EP tube. The mixture was thoroughly mixed, and the beads were separated using a magnetic stand. After washing the beads with PBS, the supernatant was discarded. In the experimental group, 350 μL of diluted primary antibody was incubated with the magnetic beads, while the control group was incubated with IgG. The incubation was carried out at 4°C with rotation for 8 hours. After cell lysis, proteins were extracted using IP lysis buffer, and their concentrations were determined. Equal amounts of protein were incubated with the magnetic beads for 8 hours. The beads were washed 5–6 times to remove nonspecific proteins. After the final wash, the beads were transferred to a new microcentrifuge tube, and 40 μL of 1×loading buffer was added. The samples were heated at 100°C for 10 minutes to denature the proteins. The denatured samples were separated by SDS-PAGE and analyzed by Western blot to detect the expression of the target protein.

ChIP was performed using the ChIP kit (#P2080S, Beyotime, Shanghai, China) according to the manufacturer’s manual. In brief, ells were crosslinked with 3.7% formaldehyde, and crosslinking was terminated with glycine. After washing with PBS, cells were lysed in SDS Lysis Buffer containing protease inhibitors and incubated on ice. Chromatin was then fragmented to 200–1000 bp by sonication, and the shearing efficiency was checked by agarose gel electrophoresis. After centrifugation to remove the pellet, the supernatant was collected and diluted with ChIP Dilution Buffer. The sample was incubated with antibody targeting designed antigen, followed by immunoprecipitation using Protein A/G magnetic beads to enrich DNA fragments bound to the target protein. The immunocomplexes underwent a series of stringent washing steps to remove nonspecific binding. Finally, the target DNA was eluted using elution buffer, and crosslinking was reversed under high-temperature conditions. The DNA was then extracted and analyzed by RT-qPCR. The primers used in this study are listed in Supplementary Table S1 .

The animal protocols were approved by the Biomedical Ethics Committee, Subcommittee on Laboratory Animal Welfare, Peking University (PUIRB-LA2022626). All mice were purchased from Vital River Laboratory Animal Technology Co. (Beijing, China) and were subcutaneously implanted with Hepa1–6 cells at a concentration of 5 × 10⁵ cells per site in 6-week-old BALB/c nude mice. Once tumors reached approximately 50 mm³ in volume, the mice were randomly divided into four groups (six mice per group) for drug administration. Tumor volume and body weight were measured daily throughout the treatment period. Tumor volume was calculated using the formula: volume (mm³) = L × W² × 0.5 (where L is the longest diameter and W is the shortest diameter). At the end of the treatment, mice were euthanized and tumors were harvested for further analysis.

RNA sequencing data have been deposited at Genome Sequence Archive for Human HRA006499 and are publicly available as of the date of this publication.

¹H and ¹³C Nuclear Magnetic Resonance (NMR) spectra were recorded on a Bruker AM-400 MHz spectrometer using C₂D₆OS (Deuterated Dimethyl Sulfoxide (DMSO)-d₆) as the solvent and tetramethyl silane (TMS) as the internal standard. Chemical shifts (δ) were reported in parts per million (ppm), and coupling constants (J) were expressed in hertz (Hz). NMR spectroscopy was used for structural elucidation of the compounds, and the detailed spectral data are shown in Supplementary Table S2 and Supplementary Figures S1A, B .

High-resolution mass spectra were obtained using an Agilent Liquid Chromatography/Mass Selective Detector Time-of-Flight (LC/MSD TOF) mass spectrometer. The mass spectrometric analysis provided accurate molecular weight information, which was used to confirm the molecular formula of the isolated compounds. The detailed High-Resolution Electrospray Ionization Mass Spectrometry (HRESIMS) data are presented in Supplementary Table S2 and Supplementary Figure S1C .

High-Performance Liquid Chromatography (HPLC) analysis was carried out using an Agilent 1260 instrument equipped with a Gemini-NX C18 110A column (4.6 × 250 mm, 5 μm). The elution was performed at a flow rate of 1 mL/min with a gradient from 5% mixed solvent (99.5% acetonitrile + 0.5% triethylamine in water) to 100% mixed solvent over 20 minutes, followed by 5 minutes at 100% mixed solvent. HPLC was used to assess the purity of the compounds, and the results are summarized in Supplementary Table S3 and Supplementary Figure S1D .

All experimental data were analyzed using GraphPad Prism 5 software. The results are presented as mean ± standard error of the mean (Mean ± SEM). For comparisons involving only two groups, a two-tailed Student’s t-test was used to calculate the p-value. When comparing more than two groups, one-way analysis of variance (One-Way ANOVA) was applied to calculate the p-value. A p-value less than 0.05 was considered statistically significant. The significant differences in the results are indicated with an asterisk: *p < 0.05.

To identify small molecules with potential inhibitory effects on CSCs stemness, we performed a primary screen using our in-house library containing 34 candidate compounds (5 μM) using sphere formation assay in Huh7 HCC cells. Among all compounds detected, compound 31 (C504244) exhibited the most potent suppression ability of sphere formation ( Figure 1A ; Supplementary Figure S2 ). Structurally, C504244 features a diaryl pyrimidine guanidine scaffold ( Figure 1B ). Its physicochemical properties were computationally evaluated using ADMETlab 3.0 (https://admetlab3.scbdd.com), and the results ( Supplementary Table S4 ) indicate favorable ADMET parameters, supporting further investigation. To further validate the inhibitory effect of C504244 on CSCs stemness, we treated two HCC cell lines, Huh7 and SK-Hep1, with C504244 at indicated concentrations ( Figure 1C ). As the data shown in Figure 1C , C504244 exhibited strong suppression effects in a dosage-dependent manner in both cell lines, suggesting a robust and consistent inhibitory effect on CSCs properties. We further assessed the potential cytotoxicity of C504244 in four normal human cells, HUVEC (Human Umbilical Vein Endothelial Cells), HDF (Human Dermal Fibroblast), WI-38 (human embryonic lung fibroblast), and PBMC (Peripheral Blood Mononuclear Cell). As shown in Supplementary Figure S3 , C504244 exhibited markedly lower toxicity in all four normal cells than in 2 HCC cells (Huh7 and SK-Hep1), supporting its tumor-selective activity.

We ALDH activity assay, CSCs marker CD24 flow cytometry analysis, and western blotting/qPCR analysis of CSCs markers, to further validate the inhibitory effect of C504244 on CSCs stemness. The results of both Huh7 and SK-Hep1 cells showed similar trends, with C504244 significantly reducing CSCs characteristics. As the data shown in Figure 2 , compared to the control group, C504244 treatments significantly decreased the proportion of ALDH+ and CD24+ cells, further confirming its inhibitory effect on CSCs characteristics. Meanwhile, the expression of several well-known CSCs markers, such as Nanog, OCT4, Sox2, and Sox9, were noticeably suppressed in C504244 treated HCC cells at both protein and mRNA levels ( Figures 2C, D ). Additionally, C504244 also suppressed the mRNA expression of ALDH ( Figure 2D ), which might contribute to the reduced activity of ALDH in C504244-treated cells.

To further evaluate the CSC-targeting effects of C504244, we sorted Huh7 cells into CD133⁺ (HCC stem cells) and CD133^- (non-stem cells) subpopulations (26–28) and treated them with vehicle control or C504244 at indicated dosages. As shown in Figures 2E–H , CD133⁺ cells were markedly more sensitive to C504244 treatment, with an IC50 of 1.927 μM, compared to 10.79 μM in CD133^- cells ( Figure 2F ). Consistently, C504244 inhibited CD133⁺ cell growth more severely than CD133^- cells ( Figure 2G ). Also, sphere formation ability was significantly reduced in CD133⁺ populations upon C504244 treatment, with no significant effects in CD133^- cells ( Figure 2H ). Taken together, these results further implicated the selective inhibitory effects of C504244 on CSC-like subpopulations in HCC.

In order to detected the effects of C504244 on HCC malignant progression, we first checked the role of it on cell viability, and found C504244 suppressed Huh7 and SK-Hep1 cell survival with IC₅₀ as 4.159 µM and 6.315 µM, respectively ( Figure 3A ). Since both decreased cell growth and increased cell death contribute to suppressed cell viability, we first evaluated the effect of C504244 on cell proliferation using both growth curve analysis and colony formation assays. As shown in Figures 3B, C , treatment with C504244 significantly inhibited cell proliferation and colony formation in both Huh7 and SK-Hep1 cells, compared to the control group, confirming its inhibitory effect on HCC cell growth. We also checked cell apoptosis using Annexin V flow cytometry analysis, and found C504244 did not induce HCC cell apoptosis significantly (data not shown), indicating C50244 induced cell loss might predominantly cause by cell proliferation inhibition. We also examined the effects of C504244 on cell migration and invasion, the key characteristics of malignant progression. Transwell migration and Matrigel invasion assays ( Figure 3D ) revealed C504244 treatment significantly reduced both migration and invasion abilities of Huh7 and SK-Hep1 cells. Additionally, the wound healing assay ( Figure 3E ) showed impaired wound closure in C504244 treated cells, indicating slowed migration. In summary, C504244 effectively inhibits HCC cell proliferation, migration, and invasion, highlighting its potential as a therapeutic agent for targeting HCC progression.

To elucidate the molecular mechanisms underlying the effects of C504244 in suppressing CSCs maintenance and malignant progression in HCC, we performed RNA sequencing using C504244-treated Huh7 cells, followed by Gene Ontology (GO) analysis. Among all the biological pathways affected by C504244, the Wnt signaling pathway ( Figure 4A ) particularly attracted our attention because the Wnt/β-catenin pathway is a central regulator of CSCs self-renewal and malignant progression in HCC. Activation of this pathway stabilizes β-catenin, facilitating its nuclear translocation and interaction with TCF4, which drives the transcription of downstream targets such as Cyclin D1 and c-Myc (17, 18, 29). To validate the sequencing results, we examined several downstream target genes of the Wnt pathway. Consistent with the RNA sequencing data, treatment with C504244 significantly inhibited the expression of 2 classic targets of Wnt signaling, c-Myc and CyclinD1 ( Figure 4B ). Unexpectedly, C504244 treatments did not alter total β-catenin levels or its phosphorylation at Ser33/37/Thr41 (which is targeted by GSK-3β for proteasomal degradation) ( Figure 4B ), nor the phosphorylation of GSK-3β (Ser9), a key kinase regulating β-catenin stability ( Figure 4B ), ruling out the possibility of upstream kinase modulation. Moreover, immunofluorescence staining confirmed that the nuclear-cytoplasmic distribution of β-catenin was unchanged by C504244 treatments ( Figure 4C ). Given that the stability and localization of β-catenin were unaltered, we hypothesized that C504244 might interfere with its transcriptional activity. To test this possibility, we performed ChIP-qPCR to assess the binding of β-catenin/TCF complex to the promoters of its target genes. C504244 treatment led to a marked reduction in TCF4 occupancy at the c-Myc and Cyclin D1 promoter regions ( Figure 4D ), suggesting transcriptional repression. We further demonstrated that C504244 significantly impaired the formation of the β-catenin/TCF4 complex ( Figure 4E ), indicating that such compound disrupts their physical interaction. To further validate the inhibitory effect of C504244 on Wnt/β-catenin transcriptional activity, we performed qRT-PCR to assess the expression levels of key target genes. Consistent with previous results, treatment with C504244 significantly decreased c-Myc and Cyclin D1 mRNA expression ( Figure 4F ). These data provide evidences suggesting that C504244 represses Wnt signaling likely through inhibiting β-catenin/TCF4 interaction, thereby impairing the transcription of key oncogenic targets critical for HCC progression.

To determine whether the anti-tumor effects of C504244 are mainly dependent on β-catenin signaling or not, we checked the effects of C504244 on HCC cells under β-catenin knockdown condition in Huh7 cells. In consistent with our previous results, we found C504244 alone significantly reduced the ALDH+ cell population, while β-catenin depletion ( Supplementary Figure S4A ) could slightly further enhance this reduction ( Supplementary Figure S4B ). We also found C504244 treatment markedly inhibited cell proliferation and migration, while β-catenin knockdown did not further suppress these phenotypes in HCC cells ( Supplementary Figures S4C, D ), indicating β-catenin plays vital roles in mediating C504244’s functions in HCC cells.

Multiple studies have confirmed that the Wnt/β-catenin signaling pathway is frequently aberrantly activated in HCC, contributing to disease progression and therapeutic resistance through various mechanisms (30–32). One critical mechanism is its role in maintaining CSCs stemness, which drives tumor resistance to anti-cancer therapies (22, 23). Consistently, analysis of our HCC organoids (with paired clinical samples) database (33) revealed significantly higher Wnt signaling activity in tumor tissues/organoids compared to adjacent normal liver tissues/organoids ( Supplementary Figure S5A ). Moreover, Wnt signaling is positively correlated with CSCs characteristics in HCC organoids ( Supplementary Figure S5B ), reinforcing its role in sustaining cancer stemness.

CSCs have been represented as the major source of therapy resistance (10, 12). Resistance to Lenvatinib, a first-line treatment for advanced HCC, has severely restricted the clinical benefits of this drug. Utilizing our HCC organoids drug-sensitivity database (33), we analyzed GSVA (Gene Set Variation Analysis) score of Wnt signaling in Lenvatinib resistant organoids compared to sensitive ones, the results revealed significantly higher Wnt pathway activation in the resistant organoids ( Figure 5A ). Furthermore, GSEA (Gene Set Enrichment Analysis) confirmed the enrichment of β-catenin target genes in the resistant organoids ( Figure 5B ). These findings suggest that aberrant Wnt activation may contribute to the development of Lenvatinib resistance.

Given these findings, we hypothesized that inhibiting Wnt/β-catenin signaling pathway could enhance the therapeutic efficacy of Lenvatinib. To test this possibility, we utilized Lenvatinib-resistant SK-Hep1 cells to assess whether combining C504244 with Lenvatinib could improve treatment response (8, 34, 35). Indeed, C504244 treatment significantly sensitize SK-Hep1 cells to Lenvatinib, with the synergy score (calculated using Synergy Finder 2.0) of 17 ( Figure 5C ), indicating a strong synergistic effect. Cell growth curve and colony formation assays confirmed the synergistic effects of C504244 and Lenvatinib ( Figures 5D, E ; Supplementary Figure S6A ). Similarly, migration and wound healing assays showed that the combination treatment effectively suppressed cell migration ( Figures 5F, G ; Supplementary Figures S6B, C ), indicating that C504244 enhances the sensitivity of Lenvatinib-resistant cells to Lenvatinib. We confirmed the effects of C504244 and Lenvatinib on HCC cells by checking the activation status of their target signaling pathways, including phosphorylation of VEGFR/Epidermal Growth Factor Receptor (EGFR) and expression of c-Myc and Cyclin D1.

Lenvatinib inhibits VEGFR/EGFR signaling in Lenvatinib sensitive Huh7 cells as expected, and C504244 suppresses β-catenin signaling, thus combined application of both drugs blocks both VEGFR/EGFR and β-catenin signaling, which contributes the synergistic effects of these drugs in Huh7 cells ( Supplementary Figure S7 ). While in Figure 5H , Lenvatinib failed to inhibit VEGFR/EGFR signalings in Lenvatinib resistant HCC cells, while C504244 suppresses β-catenin signaling, which meanwhile contributes to decreased EGFR activation (36), which might explain the synergistic effects of these drugs in HCC cells.

To further confirm the synergistic anti-tumor effect of Lenvatinib and C504244, we employed the Hep1–6 cell line, which is also resistant to Lenvatinib (35). Similar to the results observed in SK-Hep1 cells, the combination of Lenvatinib and C504244 exhibited a strong synergistic effect in Hep1–6 cells ( Supplementary Figure S8A ). Consistently, the combination treatment also significantly inhibited Hep1–6 cell proliferation, colony formation, and migration compared to either treatment alone ( Supplementary Figures S8B-D ).

To assess the therapeutic potential of combining Lenvatinib and C504244 in vivo, nude mice were implanted with Lenvatinib-resistant murine Hep1–6 tumor cells and treated with DMSO, Lenvatinib (4 mg/kg), C504244 (25 mg/kg), or the combination of both agents (35). Tumor volume was monitored over the treatment period, and the results demonstrated a significant reduction in tumor growth in the combination treatment group compared to the single-agent treatment groups ( Figure 6A ), without affecting the body weight, indicating that combination of Lenvatinib and C504244 did not cause obvious toxicity ( Figure 6B ). Moreover, Hematoxylin and Eosin (H&E) staining of major organs (heart, liver, kidney, spleen, and lung) showed no evident tissue damage, inflammation, or necrosis in mice treated with C504244 alone or in combination with Lenvatinib, further supporting the safety of C504244 and the combination regimen at the tested doses ( Supplementary Figure S9 ). Tumor masses were weighed at the end of the treatment period, and the data revealed a significant decrease in tumor weight in the combination group compared to the individual treatment groups ( Figures 6C, D ). Meanwhile, we noticed that the Ki67 and c-Myc positive cells were severely reduced in the combination group, which further confirmed C504244, in combination with Lenvatinib, exhibits synergistic effects and can reverse Lenvatinib resistance in liver cancer cells. These results collectively demonstrate that the combination of Lenvatinib and C504244 effectively inhibits tumor growth and reduces tumor weight in vivo, supporting the potential of this combination therapy for enhanced anti-tumor efficacy.