Noncanonical EZH2 activity cooperates with FOXM1 to drive tumorigenesis and advanced progression in HCC

Enhancer of Zeste Homolog 2 (EZH2), a core component of the Polycomb complex, is frequently overexpressed in hepatocellular carcinoma (HCC) and associated with aggressive tumor biology [1,2]. Increasing evidence suggests that EZH2 exerts oncogenic functions beyond its canonical H3K27me3-dependent repression, which typically promotes tumor aggressiveness through the epigenetic regulation of cell-cycle, proliferative, and cell-adhesion pathways [3,4]. Although non-canonical, H3K27me3-independent co-activator functions of EZH2 have been reported to play critical roles in cancer progression across various malignancies (e.g., breast, colon, prostate, adrenal, and hematologic cancers), yet remain poorly understood in HCC [5-8]. In this study, we sought to (i) determine whether EZH2 activates oncogenic transcriptional programs independently of H3K27me3, (ii) identify the transcription factor driving this EZH2-dependent program, and (iii) evaluate whether simultaneous inhibition of EZH2 and this effector provides therapeutic benefit in vivo.

We employed a multi-layered integrative design that sequentially incorporated transcriptomic, epigenomic, and single-cell analyses, followed by in vivo therapeutic validation (Fig. 1A). Tumor tissues from 67 HCC patients were examined by immunohistochemical staining and stratified into EZH2-High and EZH2-Low groups (IRB-approved, Supplementary Table 1), and RNA-seq was performed to define EZH2-associated transcriptional programs (GSE312183 and GSE312184). To determine whether these programs reflected canonical PRC2 function or noncanonical EZH2 activity, promoter H3K27me3 levels in EZH2-regulated clusters were assessed using ChIP-seq datasets (GSE53627) [9]. Transcription factor enrichment and integration with FOXM1 ChIP-seq datasets (GSE169998) were then applied to identify regulators of the EZH2-activated program, and their cell-type–specific distribution was evaluated using single-cell RNA-seq datasets (GSE149614) from HCC tumors to map EZH2–FOXM1 activity across distinct hepatocyte subpopulations [10,11]. Finally, to assess functional and therapeutic relevance, a model of hepatic tumor growth was generated by splenic injection of SKHEP-1 cells (IACUC-approved)–selected as a well-established model for aggressive, metastatic liver cancer–and mice were treated for 14 days with an EZH2 inhibitor (GSK126), a FOXM1 inhibitor (FDI-6), or both [12-14]. Tumor burden, histology, immunohistochemistry, qRT-PCR, and immunoblotting were performed to evaluate molecular and phenotypic responses. All human and animal studies were approved by the appropriate ethics committees.

Figure 1.

Integrated multi-omics analysis identifies a non-canonical EZH2-FOXM1 transcriptional axis driving metastatic HCC and reveals a synergistic therapeutic vulnerability. (A) Overview of the study design integrating patient stratification by EZH2 level, RNA-seq analysis, epigenome evaluation, single-cell RNA-seq mapping, and in vivo therapeutic validation. (B) Representative immunohistochemistry of EZH2 and H3K27me3 in EZH2-High and EZH2-Low HCC patients, with correlation analysis showing a weak association between EZH2 abundance and H3K27me3 levels. (C) Transcriptomic comparison between EZH2-High patients and EZH2-silenced HepG2 HCC cell line, demonstrated strong concordance, where genes elevated in EZH2-High patients showed reduced expression following EZH2 knockdown, confirming the presence of a coherent EZH2-dependent transcriptional module. GO term numbers (1-6) correspond to the top-ranked GO terms based on -Log10(P-value), as shown in Supplementary Fig. 2B. (D) Transcription factor enrichment identifying FOXM1 as the dominant regulator of EZH2-associated genes. Integration with FOXM1 ChIP-seq datasets showed strong FOXM1 occupancy at promoter regions of EZH2-activated mitotic targets, consistent with active chromatin states. (E) TCGA-LIHC analysis demonstrating a strong correlation between EZH2 and FOXM1 expression and showing that high expression of each, particularly the combined high-expression group, is associated with markedly poorer survival outcomes. (F) scRNA-seq mapping EZH2-FOXM1 targets across HCC tumors showing selective enrichment within pro-metastatic hepatocyte clusters characterized by epithelial-mesenchymal transition and migratory signatures. (G) Endogenous Co-IP of EZH2 and FOXM1 in Hep3B cells. GSK126 treatment physically disrupts the EZH2-FOXM1 complex. (H) Dual inhibition of EZH2 (GSK126, 20 mg/kg) and FOXM1 (FDI-6, 10 mg/kg) in a mouse model of hepatic tumor growth markedly reduces tumor burden and suppresses EZH2-FOXM1-regulatory targets at both RNA and protein levels, demonstrating synergistic therapeutic efficacy. Statistical significance is indicated as follows: *P<0.05, **P<0.01, ***P<0.001.

Patient tumors displayed wide variation in EZH2 expression, and immunohistochemistry demonstrated only a weak correlation between EZH2 abundance and H3K-27me3 levels, suggesting that EZH2-driven oncogenic features may arise independently of canonical PRC2-mediated repression (Fig. 1B). Consistently, transcriptomic comparison between EZH2-high patient tumors and EZH2-silenced HCC cell line demonstrated a strong concordance: genes upregulated in EZH2-High patients showed clear reduction following siRNA-mediated EZH2 knockdown in HepG2 cells (Fig. 1C and Supplementary Fig. 1). This integrative, two-layer validation indicates that the genes jointly identified across patient samples and perturbation models represent bona fide EZH2-dependent targets. The enrichment of these targets in mitotic and cell cycle– related pathways including chromosomal segregation, spindle assembly, and G2/M progression further supports their functional coherence (Supplementary Fig. 2). To determine whether this differential expression reflected canonical PRC2-mediated repression, we examined promoter H3K27me3 occupancy using ChIP-seq datasets (Supplementary Fig. 3). H3K27me3 levels remained unchanged in EZH1/2 double-knockout tissues, indicating that the EZH2-associated transcriptional program operates independently of its methyltransferase function and reflects a noncanonical regulatory role.

To identify key regulators of this EZH2-driven program, transcription factor enrichment analysis was performed, revealing FOXM1 as the dominant regulatory candidate (Fig. 1D). FOXM1 is well established as a master transcriptional driver of cell proliferation, invasion, and metastasis in various cancers, supporting its potential role in mediating EZH2-driven malignancy [15]. Integration of FOXM1 ChIP-seq datasets demonstrated strong FOXM1 occupancy at the promoter regions of key EZH2-activated mitotic genes, all of which exhibited active chromatin features (Supplementary Fig. 4). These FOXM1-bound targets were consistently upregulated in EZH2-High patients, showing expression patterns tightly aligned with EZH2 levels and indicating coordinated activation of the EZH2–FOXM1 transcriptional module (Supplementary Table 2).

In addition, analysis of TCGA-LIHC further supported these findings, showing a strong correlation between EZH2 and FOXM1 expression (R=0.86, P=5.4×10^-111), confirming the robustness of the EZH2–FOXM1 association in an independent patient cohort (Fig. 1E). Survival analyses showed that high expression of EZH2 or FOXM1 alone was associated with worse clinical outcomes (Supplementary Fig. 5), whereas combined high expression of both conferred the poorest survival, underscoring the biological and prognostic significance of this transcriptional circuit. These multi-omics results collectively indicate that EZH2 activates transcription not through H3K27me3-mediated repression but through engagement of a FOXM1-centered transcriptional network.

Next, to determine the cellular context in which this noncanonical regulatory axis operates, single-cell RNA-seq datasets from HCC tumors were examined. FOXM1-regulated EZH2 targets were enriched specifically in hepatocyte-like tumor populations characterized by markers of epithelial–mesenchymal transition, cytoskeletal remodeling, and migratory phenotypes, indicating that the EZH2– FOXM1 program is preferentially active in pro-metastatic hepatocyte subsets (Fig. 1F). Key targets such as PLK1 were localized almost exclusively to these pro-metastatic hepatocytes, rather than to proliferative but non-malignant hepatocytes.

To investigate the physical association between EZH2 and FOXM1, we performed endogenous co-immunoprecipitation (Co-IP) assays in HCC cell lines (Fig. 1G and Supplementary Fig 6). Our results demonstrated a robust and consistent physical interaction between EZH2 and FOXM1. Furthermore, treatment with the EZH2 inhibitor (GSK126) significantly disrupted this EZH2-FOXM1 complex. Collectively, this evidence substantiates that EZH2 and FOXM1 form a direct functional complex that is essential for the activation of their oncogenic target genes.

To evaluate therapeutic relevance, a model of hepatic tumor growth was generated via splenic injection of SKHEP-1 cells, which maintain sufficiently high EZH2 levels to drive enhanced proliferation and migration (Supplementary Fig. 7), and mice were treated with an EZH2 inhibitor (GSK126), a FOXM1 inhibitor (FDI-6), or both agents (Fig. 1H). Monotherapy with either inhibitor modestly reduced tumor burden, while dual inhibition resulted in a marked suppression of hepatic tumor burden, substantial reduction of tumor nodules, and a strong decrease in Ki-67 positivity without inducing weight loss or toxicity (Supplementary Fig. 8). Molecular analyses revealed pronounced suppression of core FOXM1-driven mitotic regulators—including PLK1, Cyclin F, Survivin, and UBE2C—at both the transcript and protein levels under combination treatment, with effects far exceeding those observed with either monotherapy (Supplementary Figs. 9, 10). Histological assessment demonstrated reduced mitotic activity and greater tumor differentiation in the dual-treatment group. Together, these findings show that pharmacologic disruption of the EZH2–FOXM1 axis produces strong synergistic antitumor activity, consistent with the cooperative role of these factors in driving tumor progression transcriptional programs.

This study identifies FOXM1 as the central transcriptional mediator of noncanonical EZH2 activity in HCC. Supported by direct biochemical interaction, EZH2 forms a functional complex with FOXM1 to activate a mitotic program independently of H3K27me3. This axis is selectively enriched in hepatocyte-like tumor populations that promote tumorigenic outgrowth. Dual inhibition of EZH2 and FOXM1 leads to synergistic suppression of hepatic tumor growth in vivo, demonstrating the therapeutic potential of co-targeting this regulatory circuit. Collectively, our findings extend the understanding of EZH2 biology beyond its canonical epigenetic function and uncover a druggable vulnerability in aggressive HCC. Targeting the EZH2–FOXM1 axis may provide a mechanistically grounded strategy to inhibit HCC proliferation and progression.

Abbreviations

Co-IP

co-immunoprecipitation

EZH2

Enhancer of Zeste Homolog 2

HCC

hepatocellular carcinoma

SUPPLEMENTARY MATERIAL

Supplementary material is available at Clinical and Molecular Hepatology website (http://www.e-cmh.org).

Supplementary Figure 1.

RNA-seq analysis reveals EZH2-associated transcriptional changes in HCC patients and HepG2 cells. (A) RNA-seq analysis of HCC patient cohorts (Normal, EZH2-Low, and EZH2-High) showing a stepwise increase in EZH2 expression (box plot). Hierarchical clustering of DEGs altered in HCC compared with Normal is presented as heatmap and boxplots, identifying six distinct gene clusters with characteristic expression patterns across the three groups. C1 and C2 show positive correlation with EZH2 expression, whereas C5 and C6 display negative correlation. Red squares indicate the mean expression values for each group. FC, fold change. (B) RNA-seq analysis of HepG2 cells (Control and si-EZH2) showing reduced EZH2 expression following knockdown (boxplot). DEGs regulated by EZH2 are visualized in a heatmap to illustrate transcriptional differences between the groups. Statistical significance is indicated as follows: *P<0.05, **P<0.01, ***P<0.001.

Supplementary Figure 2.

Identification of EZH2 target genes by integrative analysis of HCC patient cohorts and HepG2 datasets. (A) Scatter plot comparing fold-change values of genes between HCC patient cohorts (x-axis) and HepG2 cells treated with si-EZH2 (y-axis). EZH2-activated genes were defined as those upregulated in HCC patient cohorts and downregulated upon EZH2 knockdown in HepG2 cells, whereas EZH2-repressed genes exhibited the opposite direction of change. (B) Gene ontology (GO) enrichment analysis of EZH2-activated genes. The numbers (1–6) indicate the top enriched GO biological process terms ranked by -Log10(P-value), highlighting biological pathways functionally associated with EZH2 activation.

Supplementary Figure 3.

Integrative epigenetic analysis demonstrates that EZH2-regulated genes are independent of H3K-27me3-mediated repression. Density plots showing promoter H3K27me3 ChIP-seq signal profiles for EZH2-regulated gene sets selected from RNA-seq analysis of HCC patient cohorts, suggesting that EZH2-associated transcriptional changes occur independently of canonical H3K27me3-mediated regulation. TSS, transcription start site; WT, wild type; DKO, double knockout.

Supplementary Figure 4.

FOXM1 defines a subset of EZH2-activated targets with coordinated transcriptional activity. (A) ChIP-seq data showing FOXM1 binding at promoter regions of EZH2-activated targets, visualized as heatmap and density plot. IGV snapshots display representative loci (e.g., CCNA2, MKI67), demonstrating co-enrichment of FOXM1, H3K27ac, and H3K4me3 in HepG2 cells, supporting an active chromatin state associated with FOXM1 binding. (B) Protein-protein interaction (PPI) network of FOXM1-regulated EZH2 targets illustrating their functional interconnectivity. Dashed lines represent edges connecting distinct clusters.

Supplementary Figure 5.

Positive association between EZH2 and FOXM1 correlates with poor prognosis in TCGA-LIHC. (A) Dot plot illustrating the positive correlation between EZH2 expression and the expression of EZH2-activated target genes across samples in the TCGA-LIHC cohort. The red dashed line indicates the significance threshold (P-value=0.05). (B) Kaplan–Meier survival curves showing overall survival of TCGA-LIHC patients stratified by EZH2 expression and FOXM1 expression.

Supplementary Figure 6.

Validation of EZH2-FOXM1 physical interaction and its pharmacological disruption. (A) Endogenous Co-IP of EZH2 and FOXM1 in HepG2 cells. (B) Endogenous Co-IP in SK-HEP-1 cells demonstrating EZH2-FOXM1 interaction and its significant disruption following treatment with the EZH2 inhibitor GSK126.

Supplementary Figure 7.

Functional validation of EZH2 in HCC cell lines. (A) SK-HEP-1 and (B) HepG2 cells transfected with Control, si-NC, or si-EZH2 were analyzed for EZH2 protein expression by Western blot, confirming effective knockdown. Functional assays (proliferation, migration, and invasion) demonstrate that EZH2 silencing significantly suppresses malignant phenotypes compared with control groups. (C) In Hep3B cells, EZH2 knockdown reduces the EZH2-FOXM1 axis signature at mRNA and protein levels, while global H3K27me3 remains unchanged. Statistical significance is indicated as follows: *P<0.05, **P<0.01, ***P<0.001.

Supplementary Figure 8.

Combined inhibition of EZH2 and FOXM1 synergistically reduced tumor burden in vivo. (A) Body weight changes of mice assigned to four treatment groups (Control, GSK126, FDI-6, GSK126+FDI-6) during the experimental period. (B) Representative H& E-stained liver sections at Day 21 (scale bar=100 μm), demonstrating reduced tumor burden upon combined treatment. (C) Representative immunohistochemistry (IHC) images showing EZH2, FOXM1, and Ki-67 expression levels in liver tissues. Quantification of positively stained area (%) is shown as bar plots, indicating significantly greater suppression of tumor proliferation markers in the combination-treatment group. Statistical significance is indicated as follows: *P<0.05, **P<0.01, ***P<0.001.

Supplementary Figure 9.

Combination inhibition of EZH2 and FOXM1 suppresses EZH2-FOXM1 target gene expression in human HCC cell lines. IC50 curves and Combination Index (CI) plots for GSK126 and FDI-6 in (A) SK-HEP-1 and (B) HepG2 cells, demonstrating synergistic inhibition effects upon dual treatment. qRT-PCR analysis of FOXM1-regulated EZH2 targets showing reduced mRNA expression following combination treatment in (C) SK-HEP-1 and (D) HepG2 cells. Western blot analysis confirms suppression of protein expression levels for these targets in (E) SK-HEP-1 and (F) HepG2 cells. Statistical significance is indicated as follows: *P<0.05, **P<0.01, ***P<0.001.

Supplementary Figure 10.

Effects of EZH2 and FOXM1 inhibition on EZH2-FOXM1 target gene expression in mouse in vivo model. (A) qRT-PCR analysis of FOXM1-regulated EZH2 targets expression in liver tissues from mice injected with vehicle control, GSK126 (EZH2 inhibitor), FDI-6 (FOXM1 inhibitor), or the combination of GSK126 and FDI-6. (B) Western blot analysis showing protein levels of the corresponding target genes in mouse liver tissues under the same groups. Combination treatment resulted in a synergistic suppression of target gene expression compared with single-agent injections. Statistical significance is indicated as follows: *P<0.05, **P<0.01, ***P<0.001.

Supplementary Table 1.

Disease characteristics of HCC patients

Supplementary Table 2.

Expression count table of 58 FOXM1-regulated EZH2 target genes

Supplementary Table 3.

Primers

Supplementary Table 4.

Antibodies