CHD4 epigenetically coordinates genomic instability and immunosuppression to drive pan-cancer progression and confer HDAC inhibitor sensitivity

Abstract

Chromodomain Helicase DNA-Binding Protein 4 (CHD4), the core ATPase subunit of the nucleosome remodeling and deacetylation (NuRD) complex, is a key epigenetic regulator. However, a systematic pan-cancer perspective on its functions, particularly its coordinated regulation of genomic stability alongside the tumor immune microenvironment, remains lacking. This study performed an integrated multi-omics analysis using data from The Cancer Genome Atlas (TCGA) and complementary genomic databases. This included systematic profiling of CHD4 expression, genomic alterations, and clinical associations across cancers. We investigated its correlations with markers of genomic instability, immune cell infiltration, and therapy response. Functional enrichment and pharmacogenomic analyses were conducted, supported by in vitro validation in osteosarcoma models. CHD4 was frequently upregulated across multiple cancer types, and its elevated expression was associated with poorer patient prognosis in several malignancies. Pan-cancer analysis revealed that high CHD4 expression correlated significantly with markers of genomic instability, such as homologous recombination deficiency (HRD) and loss of heterozygosity (LOH), and concurrently with an immunosuppressive tumor microenvironment—characterized by reduced CD8 + T cell infiltration and elevated expression of immune checkpoint molecules. Mechanistically, CHD4 expression was closely linked to core components of the NuRD complex, including HDAC1 and HDAC2, suggesting its involvement in chromatin compaction and transcriptional regulation associated with these phenotypes. Furthermore, tumors exhibiting high CHD4 expression showed increased sensitivity to histone deacetylase (HDAC) inhibitors, including vorinostat and panobinostat. This study establishes CHD4 as a pan-cancer epigenetic regulator whose expression is linked to both genomic instability and immune suppression. Furthermore, CHD4 shows promise as a predictive biomarker for sensitivity to HDAC inhibitors, highlighting its potential as a biomarker for guiding epigenetics-based therapeutic strategies.

Supplementary Information

The online version contains supplementary material available at 10.1007/s10238-026-02081-y.

Introduction

Cancer remains a major global public health threat and the second leading cause of mortality worldwide. According to 2022 global cancer statistics, malignant neoplasms account for approximately 10 million annual deaths, with the disease burden continuing to rise [1, 2]. Despite considerable advances in cancer treatment—particularly the introduction of immune checkpoint inhibitors and molecularly targeted therapies, which have markedly improved survival for patients with various cancers—tumor heterogeneity and therapy resistance remain significant obstacles to improving clinical efficacy [3–5]. Tumorigenesis is a complex, multi-step process driven by genomic mutations, epigenetic modifications, and remodeling of the tumor microenvironment. Collectively, these alterations enable tumor cells to acquire sustained proliferative capacity, undergo metabolic reprogramming, and evade immune surveillance [6, 7]. Epigenetic regulation, which confers phenotypic plasticity beyond the genetic code, thus underlies these central challenges. Within the epigenetic machinery, chromatin remodeling complexes, in particular, dynamically regulate gene expression programs by altering chromatin accessibility, thereby influencing tumor cell plasticity, genomic stability, and immune recognition [8].

Among these complexes, the nucleosome remodeling and deacetylation (NuRD) complex serves as a key epigenetic modulator, coordinating transcriptional repression and DNA damage response through its ATPase and histone deacetylase activities [9, 10]. Central to the NuRD complex is CHD4 (Chromodomain Helicase DNA-Binding Protein 4), its catalytic ATPase subunit of the NuRD complex, which is implicated in diverse biological processes including DNA repair, cell cycle progression, and stemness maintenance [11–13]. Its functional role in cancer appears context-dependent: while CHD4 can maintain genomic integrity under stress conditions, it also facilitates tumor progression, metastasis, and chemoresistance in several malignancies, such as gastric, ovarian, and triple-negative breast cancers [14–16]. In line with this functional duality, the net impact of CHD4 likely depends on the precise cellular context. However, a systematic, pan-cancer analysis that can disentangle these context-specific effects is lacking. Consequently, its coordinated effects on genomic instability and the tumor immune microenvironment remain poorly defined. Furthermore, whether CHD4 expression holds predictive value for epigenetic therapies—such as histone deacetylase (HDAC) inhibitors—has not been comprehensively explored across cancer types.

This study deployed an integrated multi-omics approach to (i) systematically profile CHD4 across cancers, (ii) decipher its associations with DNA repair, immune infiltration, and genomic instability, and (iii) assess its potential as a biomarker for HDAC inhibitor response. This pan-cancer framework enables the distinction of conserved CHD4-associated pathways from context-specific effects, thereby informing its potential both as a broad-spectrum therapeutic target and as a predictive biomarker for epigenetics-guided precision oncology. An outline of our methodological approach is provided in Fig. 1.

Fig. 1.

Study flowchart

Materials and methods

Data sources and processing

Transcriptomic data (RNA‑seq in TPM format) and matched clinical information for 34 cancer types were obtained from the UCSC Xena platform (https://xenabrowser.net/), which integrates The Cancer Genome Atlas (TCGA), Therapeutically Applicable Research to Generate Effective Treatments (TARGET), and Genotype‑Tissue Expression (GTEx) datasets [17]. After removing cancer types with fewer than three samples, the final pan‑cancer cohort comprised 19,131 samples. Expression values were normalized using a log2(TPM + 0.001) transformation. CHD4 expression in cancer cell lines was acquired from the Cancer Cell Line Encyclopedia (CCLE, https://Sites.broadinstitute.org) [18]. Genomic alteration profiles (mutations, copy‑number variations) and DNA methylation data were retrieved from cBioPortal (http://www.cbioportal.org/) [19]. Protein‑level validation was performed using the Clinical Proteomic Tumor Analysis Consortium (CPTAC) dataset via UALCAN (http://ualcan.path.uab.edu/) [20]. Immunohistochemistry (IHC) images were collected from the Human Protein Atlas (HPA, https://www.proteinatlas.org/) [21]. The cancer types analyzed in this study are listed in Table S1 (Supplementary material 2).

CHD4 expression and clinical correlation analysis

Differential CHD4 mRNA expression between tumor and adjacent normal tissues was assessed using the limma R package (version 3.56.0). For cancers with at least 15 paired samples (n = 15 cancer types), paired Wilcoxon tests were applied. Protein-level validation was performed using data from the CPTAC via UALCAN. IHC images were examined to visually compare CHD4 protein expression patterns between tumor and normal tissues for six representative cancer types.

To evaluate the association between CHD4 expression and disease progression, we visualized its expression levels using boxplots stratified by TNM stage, pathological stage, and tumor grade based on TCGA clinical annotations. Patients were stratified into CHD4 “high” and “low” expression groups based on the optimal cut-off value determined by the surv_cutpoint function (survminer R package, version 0.4.9). This function identifies the expression threshold that maximizes the survival difference (assessed by log-rank statistics) between the two resultant groups across the entire cohort. Diagnostic performance was evaluated by receiver operating characteristic (ROC) curves using the pROC package (version 1.18.0), with an area under the curve (AUC) > 0.7 considered clinically informative. Survival analysis for overall survival (OS), disease‑specific survival (DSS), disease‑free interval (DFI), and progression‑free interval (PFI) was performed using univariate Cox regression. Kaplan‑Meier curves were plotted, and differences between groups were compared using the log‑rank test.

Analysis of genomic alterations and genomic instability

The frequency and type of CHD4 genomic alterations (mutations, amplifications, deep deletions) were summarized using the “Cancer Types Summary” module in cBioPortal. Pan‑cancer single‑nucleotide variant (SNV) profiles were visualized with the maftools R package (version 2.16.0). The prognostic impact of CHD4 copy‑number variations (CNVs) was assessed using Kaplan‑Meier analysis within the TIDE platform [22]. Correlations between CHD4 CNVs and the activity of key oncogenic pathways (TP53, RTK/RAS, PI3K, cell cycle) were also examined.

Genomic instability metrics—including tumor mutational burden (TMB), microsatellite instability (MSI), mutant‑allele tumor heterogeneity (MATH) score, neoantigen load, tumor purity, ploidy, homologous recombination deficiency (HRD) score, loss of heterozygosity (LOH), aneuploidy score, and the ratio of non‑synonymous to synonymous mutations—were calculated using maftools. Patients were classified into four groups based on CHD4 expression and immune/genomic subtypes as described by Thorsson et al. [23], and mean values of each instability metric were visualized in a heatmap.

Analysis of DNA Repair, tumor Stemness, and epigenetic regulation

Correlations between CHD4 expression and five DNA mismatch repair (DMMR) genes [24], eleven histone‑modification‑associated proteins, and four DNA methyltransferases (DNMTs) were calculated across cancer types [25]. The association between CHD4 and homologous recombination repair (HRR) activity was evaluated using a gene signature derived from the ARIEL3 trial via GEPIA2 [26]. Tumor stemness scores—including RNA‑based scores (RNAss, EREG‑EXPss) and DNA‑methylation‑based scores (DNAss, EREG‑METHss, DMPss, ENHss)—were correlated with CHD4 expression. Relationships between CHD4 promoter methylation, cytotoxic T lymphocyte (CTL) infiltration, and patient survival were assessed using the methylation module of the TIDE platform and survival R package (version 3.6.0). Differential methylation analysis across specific genomic loci (DHS sites, CpG shores/shelves) was performed, and methylation levels were correlated with CHD4 expression. Additionally, correlations between CHD4 and 44 RNA modification regulators (involved in m1A, m5C, and m6A pathways) were visualized in a heatmap [27].

Analysis of alternative splicing events

Clinically relevant alternative splicing (AS) events of CHD4 were identified using the ClinicalAS module of the OncoSplicing platform (http://www.oncosplicing.com/) [28]. Splicing percentages (PSI values) in tumor versus normal tissues from the TCGA‑GTEx pan‑cancer dataset were visualized with PanPlot. The prognostic significance of significant AS events was evaluated via Kaplan‑Meier survival analysis in the pan‑cancer cohort.

Functional enrichment and interaction analysis

Experimentally validated protein‑protein interactions (PPIs) for CHD4 were retrieved from the STRING database (confidence score > 0.7) [29]. Somatic alteration profiles across key oncogenic pathways were examined using UALCAN, while associations between CHD4 expression and pathway activity were evaluated with GEPIA2 [30]. The top 100 genes co‑expressed with CHD4 across all cancers were identified using GEPIA2. Functional enrichment analysis of these genes for Gene Ontology (GO) terms was performed with clusterProfiler (version 4.8.0) and org.Hs.eg.db. Gene Set Enrichment Analysis (GSEA) was conducted using the Hallmark and KEGG collections on samples grouped by CHD4 expression (top vs. bottom 30%) [31]. To investigate the functional states associated with CHD4 expression, we utilized 14 cancer-related gene sets from the CancerSEA database, covering processes such as apoptosis, cell cycle, DNA damage response, EMT, hypoxia, inflammation, invasion, metabolism, proliferation, stemness, and angiogenesis [32]. Gene set variation analysis (GSVA) was performed using the R package GSVA (version 1.48.0) to calculate single-sample enrichment scores for each functional state. Scores were then Z‑normalized across samples to generate comparable pathway activity metrics. Associations between CHD4 expression and each functional score were assessed using Pearson correlation and visualized in scatter plots.

Characterization of the Pan‑Cancer immune microenvironment

A curated set of 68 immune‑related gene signatures was obtained from UCSC Xena. Stromal, immune, and ESTIMATE scores were computed using the ESTIMATE R package (version 1.0.13). Immune cell infiltration levels for 19 immune cell types and 2 stromal cell types were estimated using six deconvolution algorithms (TIMER, CIBERSORT, xCell, EPIC, MCP‑counter, quanTIseq) implemented in the IOBR package (version 2.0) [33]. We note that immune deconvolution from bulk RNA‑seq infers relative, not absolute, cell abundances and does not capture spatial architecture. The correlations reported should therefore be interpreted as associations between gene expression and computationally inferred immune cell profiles. Correlations between CHD4 expression and immune checkpoint molecules, immunomodulatory genes (chemokines, receptors, MHC molecules, inhibitors, stimulators), and cytokine/checkpoint‑blockade treatment responses (modeled via TISMO, http://tismo.cistrome.org/) were assessed [34]. CHD4 expression patterns across six immune subtypes (C1–C6) were analyzed using TISIDB (http://cis.hku.hk/TISIDB/) [35]. Spatial co‑localization of CHD4 with epithelial (CDH1) and proliferation (CCND1) markers in breast cancer was examined using spatially resolved transcriptomics data from SpatialDB (https://www.spatialomics.org/SpatialDB/) [36]. Single‑cell‑level CHD4 expression across malignancies was explored via the Tumor Immune Single‑Cell Hub (TISCH, http://tisch.comp-genomics.org/) database [37].

Evaluation of treatment response and drug sensitivity

Associations between CHD4 expression and response to chemotherapy, targeted therapy, and endocrine therapy in breast cancer (BRCA), glioblastoma (GBM), colon adenocarcinoma (COAD), and ovarian cancer (OV) were assessed using ROC Plotter (https://rocplot.org/). The Connectivity Map (CMAP, L1000 platform) was queried to identify small‑molecule compounds that reverse CHD4‑high expression signatures (connectivity score > 90) [38]. Correlations between CHD4 expression and half‑maximal growth inhibitory concentration (GI₅₀) for candidate compounds were analyzed using the COMPARE tool from the NCI Developmental Therapeutics Program. Finally, sensitivity associations between CHD4 expression and drug response were systematically validated across five independent pharmacogenomic databases: CTRP (v2.0), GDSC1, GDSC2, PRISM, and RNAactDrug.

Experimental methods

The experimental methods are detailed in Supplementary material 3.

Statistical analysis

All analyses were performed in R (version 4.2.1). Group comparisons for continuous variables used the Student’s t‑test (normal distribution) or Mann–Whitney U test (non‑normal). Survival differences were assessed with the log‑rank test. Correlation analyses employed Pearson or Spearman methods as appropriate. All analyses involving multiple comparisons across cancer types or genomic/immunological features were corrected for the false discovery rate (FDR) using the Benjamini-Hochberg method. An FDR-adjusted q-value < 0.05 was considered statistically significant for all reported conclusions.

Results

CHD4 is upregulated across multiple cancers and associates with clinical progression

Analysis of 34 cancer types from TCGA and GTEx revealed consistent upregulation of CHD4 mRNA in tumors—including BRCA, COAD, and LUAD—compared to normal tissues (Fig. 2A). This pattern was further corroborated in paired samples from 15 cancer types, where tumor tissues exhibited significantly higher CHD4 expression than matched normal samples (Fig. S1A). Widespread CHD4 expression was also observed across 32 cancer cell lines in the CCLE database, with particularly elevated levels in leukemias, lymphomas, and several solid tumors (Fig. 2B). Protein‑level upregulation was validated by CPTAC data, which showed substantial CHD4 enrichment in 10 cancer types, including BRCA, COAD, and GBM (Fig. 2C). IHC images from the HPA database further supported elevated CHD4 protein expression in tumor tissues from BRCA, COAD, and LIHC (Fig. 2D).

Fig. 2.

Pan-cancer analysis of CHD4 expression and prognostic relevance. (A) Differential CHD4 mRNA expression between tumor and normal tissues from TCGA and GTEx datasets. (B) CHD4 mRNA expression in cancer cell lines from the CCLE datasets. (C) CHD4 protein expression in pan-cancer tissues from the UALCAN-CPTAC dataset. (D) Immunohistochemistry images of CHD4 protein expression in normal and tumor tissues from the HPA database. (E) CHD4 expression across different TNM stages, histological grades, and pathological stages in pan-cancer. (F) Forest plots of hazard ratios for the association between CHD4 expression and OS, DSS, DFI, and PFI in TCGA cancers. * p < 0.05, ** p < 0.01, *** p < 0.001; ns: not significant

Clinically, elevated CHD4 expression showed a strong positive association with advanced TNM stage, pathological stage, and tumor grade across multiple cancers (Fig. 2E), suggesting a link to disease progression. In diagnostic evaluation, CHD4 expression discriminated tumor from normal tissue with AUC > 0.7 in several cancers, including BLCA, BRCA, and HNSC (Fig. S1B). Survival analysis revealed that elevated CHD4 expression was associated with poorer overall survival in LIHC, KICH, and UVM, but correlated with better outcomes in GBMLGG, THYM, and PAAD (Fig. 2F). Kaplan–Meier curves further illustrated significant survival stratification between CHD4‑high and ‑low groups in specific cancer types (Fig. S2A).

CHD4 genomic alterations are linked to markers of genomic instability

The genomic alteration landscape of CHD4 was characterized by frequently amplification in UCS, OV, TGCT, and LGG, while SNVs predominated in UCEC, UCS, SKCM, and STAD; deep deletions were rare (Fig. 3A, B). Increased CHD4 copy number was associated with improved survival in AML, LIHC, and STAD, but predicted poorer outcomes in BRCA‑LumA, BRCA‑TN, DLBCL, and COADREAD (Fig. 3C). Co‑alteration analysis frequently implicated TP53 in CHD4‑altered tumors (Fig. S3A), and CHD4 expression correlated with mutational status in key oncogenic pathways—including TP53, RTK/RAS, PI3K, and Cell Cycle (Fig. 3D).

Fig. 3.

CHD4 expression is associated with genomic instability. (A) Genetic alteration features of CHD4 in TCGA pan-cancer cohorts via cBioPortal. (B) Landscape of CHD4 SNVs across cancer types. (C) Kaplan-Meier curves from the TIDE platform showing the predictive significance of CHD4 CNVs in six cancer types. (D) Co-mutation analysis between CHD4 and key oncogenic signaling pathways. (E-G) Lollipop plots of Spearman correlation between CHD4 expression and TMB (E), MSI (F), and LOH (G). (H-J) Radar charts showing Spearman correlation between CHD4 expression and HRD (H), tumor ploidy (I), and aneuploidy score (J). The colored curve indicates the correlation coefficient; the blue value indicates the range. * p < 0.05, ** p < 0.01, *** p < 0.001; ns: not significant

CHD4 expression showed cancer‑type‑specific correlations with TMB and MSI, being positively associated in LIHC but negative correlations in STES, THYM, and MESO (Fig. 3E, F). Notably, CHD4 expression exhibited consistent, pan-cancer positive correlations with HRD and LOH across cancer types (Fig. 3G, H), underscoring its pivotal role in DNA damage repair. It was also positively associated with tumor heterogeneity (MATH score) in 9 of 13 cancers examined (Fig. S3B) and with increased tumor purity in 18 cancers (Fig. S3C). In contrast, correlations between CHD4 expression and ploidy or aneuploidy were heterogeneous across malignancies (Fig. 3I, J), indicating context-dependent relationships. Tissue‑specific relationships were observed between CHD4 expression and neoantigen load (Fig. S3D), with a stronger correlation to synonymous versus non‑synonymous mutation rates (Fig. S3E, F). Integrated heatmap analysis further connected CHD4 expression to co‑occurring immune and genomic instability phenotypes (Fig. S3G). Collectively, these results establish CHD4 as a key modulator of genomic instability in human cancers.

CHD4 associates with DNA Repair, Stemness, and epigenetic regulatory networks

Given the observed associations with genomic instability, we examined whether CHD4 expression correlates with DNA repair, stemness, and epigenetic regulation. CHD4 expression positively correlated with MMR genes across multiple cancer types, including THYM, KICH, and BRCA (Fig. 4A). Furthermore, CHD4 showed a significant positive association with HRR activity in 12 cancer types, including KICH and BRCA (Fig. 4B). Tumor stemness indices exhibited tissue‑specific correlations: DNA methylation‑based scores (DNAss, EREG‑METHss, ENHss, DMPss) correlated positively with CHD4 in HNSC and LUSC, whereas RNA‑based stemness scores (RNAss, EREG.EXPss) showed negative correlations in KIRP and THCA (Fig. S4A–F).

Fig. 4.

CHD4 is associated with DNA repair, stemness and epigenetic modifications. (A) Heatmap of Spearman correlations between CHD4 and five MMR genes across pan-cancer. (B) Scatter plot of the correlation between CHD4 expression and HRR score. (C) Heatmap of correlations between CHD4 and 11 histone modification-associated proteins. (D) Scatter plots of correlations between CHD4 and key epigenetic regulators (HDAC1, HDAC2, EP300) in three representative cancer types. (E) Heatmap of associations between CHD4 and four DNMTs. (F) Scatter plots of correlations between CHD4 promoter methylation levels and CTL marker expression. (G) Bubble plot of differential CHD4 methylation levels at specific genetic loci. (H) Bubble plot of correlations between CHD4 expression and its methylation levels across various genetic loci

CHD4 expression strongly and positively correlated with core NuRD components HDAC1 and HDAC2, and more moderately with HDAC3 and SIRT1 across cancers (Fig. 4C, D). Conversely, negative correlations were observed with histone acetyltransferases EP300 and CREBBP. At the DNA methylation level, CHD4 correlated positively with DNMTs in LUSC, READ, and UCEC, but inversely in ACC and KICH (Fig. 4E). CHD4 expression negatively correlated with its promoter methylation (Fig. S4G), which in turn predicted poorer survival in COADREAD and PAAD (Fig. S5A) and was linked to cytotoxic T lymphocyte infiltration and clinical risk (Fig. 4F).

Differential methylation analysis revealed tumor‑specific hypermethylation at DHS sites, CpG shores, and shelves in seven cancer types (Fig. 4G). Methylation levels in these regions negatively correlated with CHD4 expression (Fig. 4H). Finally, CHD4 was broadly associated with 44 RNA modification regulators involved in m1A, m5C, and m6A pathways (Fig. S5B).

Alternative splicing of CHD4 correlates with patient prognosis

Analysis of clinically relevant AS events identified 25 significant AS events in CHD4 (Table S2). We focused on the CHD4_alt_3prime_57972 event, which showed substantial variability across cancers (Fig. 5A, B). Its PSI values were elevated in KIRP and PCPG compared to normal tissues, but reduced in BRCA, CHOL, and KICH. Higher PSI of this event was associated with poorer overall survival in ESCA, KICH, KIRC, and SARC (Fig. 5C). Parallel analysis of the CHD4_AA_19897 event further supported the prognostic relevance of CHD4 splicing patterns (Fig. S6A–C).

Fig. 5.

CHD4 alternative splicing correlates with patient prognosis. (A) Reads-in, reads-out, and PSI values of the alternative splicing event CHD4_alt_3prime_57972 across pan-cancer, adjacent, and normal tissues. (B) Differences in PSI values between tumor-adjacent and tumor-normal comparisons, and their association with OS and PFI. The red dashed line indicates FDR = 0.05(Benjamini-Hochberg correction). Point size corresponds to tumor PSI values. (C) Kaplan-Meier curves from OncoSplicing showing the prognostic significance of the CHD4_alt_3prime_57972 event

Functional enrichment implicates CHD4 in DNA damage response and oncogenic pathways

Protein‑protein interaction analysis identified ten experimentally validated CHD4‑binding partners (Fig. 6A). Pathway‑based evaluation indicated elevated CHD4 expression in BRCA patients with somatic alterations in chromatin modifiers, Hippo, Nrf2, and SWI/SNF pathways, but decreased expression in PRAD (Fig. 6B). CHD4 expression consistently correlated with activation of these pathways across various cancers (Fig. 6C; Table S3).

Fig. 6.

Functional enrichment and co-expression analysis of CHD4. (A) Protein-protein interaction (PPI) network of CHD4 from the STRING database. (B) CHD4 expression in six cancer types stratified by somatic alteration status in indicated pathways. (C) Correlations between CHD4 expression and oncogenic pathway signatures. (D) Heatmap of pan-cancer correlations between CHD4 and its top five co-expressed genes. (E) GO enrichment analysis of the top 100 CHD4 co-expressed genes. (F) Scatter plots of Pearson correlation between CHD4 expression and functional state activity scores for 14 cancer-related processes

Co‑expression analysis revealed that the top 100 CHD4‑correlated genes included SF3B2, FUS, UBTF, DHX9, and DDX39B (Fig. 6D). GO enrichment analysis indicated strong involvement in homologous recombination, double‑strand break repair, chromatin assembly, and cell cycle checkpoint regulation (Fig. 6E; Table S4). GSEA showed coordinated activation of E2F targets, G2M checkpoint, glycolysis, oxidative phosphorylation, and EMT in CHD4‑high samples (Fig. S7A). Single‑cell functional analysis confirmed positive correlations between CHD4 expression and pro‑tumorigenic states, including cell cycle progression, DNA damage response, proliferation, and stemness (Fig. 6F).

CHD4 is associated with an immunosuppressive tumor microenvironment

CHD4 expression correlated positively with TGF‑β signaling and myeloid activation markers (e.g., TREM1), but negatively with antitumor immunity features, including CD8 + T cell infiltration, T cell receptor signaling, interferon response, and MHC‑I expression (Fig. S8A). CHD4 was also associated with upregulation of immune checkpoint molecules, including PD‑L1, CTLA‑4, LAG‑3, and TIGIT (Fig. S8B).

ESTIMATE‑based deconvolution revealed cancer‑type‑specific correlations between CHD4 and stromal, immune, and ESTIMATE scores, with consistently negative associations in SARC and ACC and positive associations in LGG (Fig. 7A). Cellular infiltration analysis linked high CHD4 to reduced antitumor effector cells (M1 macrophages, γδ T cells) and enrichment of immunosuppressive populations, including cancer‑associated fibroblasts, endothelial cells, M0 macrophages, neutrophils, mast cells, hematopoietic stem cells, and Tregs (Fig. 7B). Positive correlations with CD4 + and CD8 + T cells likely reflect an exhausted T cell phenotype. CHD4 expression correlated with immunomodulatory genes in a cancer‑dependent manner—positively with 89 factors in THCA but negatively in THYM (Fig. S9A). CHD4 was highest in IFN‑γ‑dominant immune subtypes (C1, C2) and lowest in the C6 subtype (Fig. 7C).

Fig. 7.

CHD4 expression is associated with a suppressive tumor immune microenvironment across multi-omic dimensions. (A) Heatmap of Spearman correlation between CHD4 expression and tumor microenvironment scores (Stromal, Immune, ESTIMATE) and 21 immune cell infiltration subsets. (B) Scatter plots of the three cancer types with the most significant correlations between CHD4 expression and Stromal, Immune, or ESTIMATE scores. (C) CHD4 expression across six immune subtypes in UCEC, HNSC, LUAD, STAD, LUSC, and LIHC. (D) Spatial transcriptomic sections of CHD4 with epithelial marker CDH1 and proliferation marker CCND1, showing spatial co-expression patterns. (E) Single-cell resolution analysis of CHD4 expression across cell types in the PAAD_GSE111672 dataset

Spatial transcriptomics confirmed CHD4 co‑localization with epithelial marker CDH1 and proliferation marker CCND1 in BRCA (Fig. 7D). Single‑cell RNA‑seq analysis revealed CHD4 expression in malignant, epithelial, endothelial, and fibroblast populations (Fig. 7E). CHD4 was upregulated following cytokine stimulation (IFN‑γ, TNF‑α) and immune checkpoint blockade (anti‑CTLA‑4/PD‑1) (Fig. S9B, C).

CHD4 expression predicts therapy response and sensitivity to HDAC Inhibition

This study subsequently assessed the clinical translational potential of CHD4 as a prognostic indicator of treatment response. ROC plotter analysis indicated that elevated CHD4 expression predicted better response to capecitabine and fluorouracil‑based therapies in COAD, but was associated with non‑response in OV (Fig. 8A). CMAP screening identified 30 potential compounds targeting CHD4‑associated vulnerabilities (Fig. 8B), with significant enrichment for HDAC and topoisomerase inhibitors in mechanism‑of‑action analysis (Fig. 8C). COMPARE‑based evaluation showed that elevated CHD4 expression was associated with reduced drug sensitivity (higher GI₅₀) in leukemia, colon cancer, and melanoma, while lower CHD4 levels predicted higher sensitivity in NSCLC, ovarian, prostate, and breast cancers (Fig. 8D).

Fig. 8.

CHD4 predicts therapy response and potential targeted drugs. (A) CHD4 expression in responders vs. non-responders and ROC curve for predicting response. (B) Heatmap of the top 30 candidate drugs for CHD4-high tumors from cMAP. (C) MoA analysis for the top 30 compounds in (B). (D) Correlation between mocetinostat GI50 and CHD4 expression in NCI60 cell lines. (E-I) Lollipop plots of Spearman correlation between CHD4 expression and drug sensitivity in five pharmacogenomic databases: (E) CTRP, (F) PRISM, (G) GDSC1, (H) PRISM, (I) RNAactDrug (computational score). (J) Identification of small-molecule compounds targeting CHD4 using the xSum algorithm. (K) Bubble plot of Spearman correlation between CHD4 expression and chemotherapy drug sensitivity across four integrated database

Integrated pharmacogenomic analysis across five independent databases consistently linked high CHD4 expression to increased sensitivity to HDAC inhibitors, as evidenced by significant negative correlations between CHD4 levels and IC₅₀/AUC values for vorinostat, panobinostat, and entinostat (Fig. 8E–I). The XSum algorithm nominated entinostat (MS‑275) as a top candidate for reversing CHD4‑driven transcriptional signatures (Fig. 8J). Expanded analysis revealed that high CHD4 expression also correlated with broad sensitization to conventional chemotherapeutic agents across four additional databases (Fig. 8K).

CHD4 promotes Proliferation, Migration, and invasion in osteosarcoma cells

CHD4 was significantly upregulated in osteosarcoma cell lines at both mRNA and protein levels compared to normal osteoblasts, with highest expression in 143B and Saos‑2 cells (Fig. 9A, B). Stable CHD4‑knockdown models were established using lentiviral RNA interference, with shCHD4‑2 showing optimal silencing efficiency (Fig. 9C, D). CHD4 depletion markedly reduced expression of metastasis‑associated proteins MMP2 and MMP9 (Fig. 9E). Functionally, CHD4 knockdown significantly suppressed cellular proliferation (Fig. 9F), impaired clonogenic survival (Fig. 9G), inhibited invasion in Transwell assays (Fig. 9H), and attenuated migration in wound healing assays (Fig. 9I). Collectively, these results establish CHD4 as a critical oncogenic driver in osteosarcoma, essential for sustaining proliferative, migratory, and invasive capacities.

Fig. 9.

CHD4 is upregulated in osteosarcoma and essential for malignant phenotypes in vitro. (A, B) CHD4 mRNA (A) and protein (B) expression in normal osteoblastic (hFOB1.19) and osteosarcoma cell lines. (C, D) Validation of CHD4 knockdown by qPCR (C) and Western blot (D) in 143B and Saos-2 cells expressing control (sh-NC) or CHD4-targeting shRNAs. (E) Western blot analysis of MMP2 and MMP9 expression following CHD4 knockdown. (F, G) The impact of CHD4 knockdown on the proliferative capacity of osteosarcoma cells was assessed by CCK-8 assay (F) and colony formation assay (G). (H) Cellular invasion capacity determined by Transwell assay (Scale bar: 100 μm). (I) Cell migration evaluated by wound healing assay (Scale bar: 100 μm). * P < 0.05, ** P < 0.01, *** P < 0.001

Discussion

This integrated pan-cancer multi-omics analysis, complemented by functional validation, elucidates the multifaceted role of the chromatin remodeler CHD4. We demonstrate that CHD4 is frequently upregulated across diverse malignancies and that its elevated expression is associated with adverse prognosis, markers of genomic instability, and an immunosuppressive tumor microenvironment. Crucially, we identify CHD4 as a novel predictive biomarker for sensitivity to HDAC inhibitors. These findings position CHD4 as a master epigenetic coordinator linking the regulation of genomic integrity, immune evasion, and therapeutic response. Despite these advances, a systematic, pan-cancer evaluation of CHD4 as a unifying epigenetic node coordinating these disparate processes, and its translational utility as a predictive biomarker, has been lacking. This work not only corroborates recent reviews highlighting CHD4 as a key player in cancer progression [9], but also significantly extends its translational relevance by defining its potential in guiding epigenetics-based precision oncology.

This study confirms that CHD4 is broadly upregulated at both mRNA and protein levels in multiple cancers, and its expression positively correlates with advanced tumor stage and grade. This aligns with its previously reported pro-tumorigenic functions in gastric, ovarian, and breast cancers [14–16]. However, the prognostic impact of CHD4 is highly context-dependent. While high CHD4 expression signifies poorer survival in cancers like LIHC and KICH, it paradoxically correlates with better outcomes in GBMLGG and THYM. This functional duality is reflected in the literature, where CHD4 can maintain genomic stability in Ewing sarcoma [39] yet promote chemoresistance in acute myeloid leukemia [40]. Importantly, these contrasting prognostic associations are not inconsistencies but rather pivotal clues highlighting the functional plasticity of CHD4 as an epigenetic node. We propose that CHD4’s functional output—and consequently its impact on patient survival—is dictated by the prevailing cellular context of a given tumor. In contexts of high genomic instability and replicative stress (e.g., certain gliomas or thymomas), the DNA damage repair function of CHD4, coordinated through the NuRD complex, may become predominant [41, 42]. Here, elevated CHD4 could help maintain baseline genomic integrity, potentially delaying disease progression and manifesting as a protective association. Conversely, in cancers with a relatively stable genome but strong immunoediting pressure (e.g., hepatocellular carcinoma), the oncogenic facets of CHD4—particularly its drive toward immunosuppression and epigenetic silencing—likely prevail, leading to its association with adverse outcomes [43]. Decoding the molecular switches (e.g., specific co-mutations like TP53 status, upstream pathway activity, or tissue-specific chromatin contexts) that dictate this functional output will be crucial for predicting CHD4’s role in individual tumors and for developing context-aware therapeutic strategies. Our functional validation in osteosarcoma cells, a cancer characterized by genomic instability, supports its tumor-promoting role in this specific context, establishing CHD4 as essential for sustaining proliferative, migratory, and invasive capacities.

Mechanistically, our data, consistent with the canonical function of the NuRD complex [11], demonstrate that CHD4 exerts its pleiotropic effects primarily through epigenetic modulation. The robust pan-cancer correlation between CHD4 expression and core NuRD components (HDAC1/2), coupled with an inverse association with histone acetyltransferases (EP300, CREBBP), strongly suggests that CHD4 serves as a central node enforcing NuRD-mediated histone deacetylation and transcriptional repression. This chromatin remodeling capacity underpins its context-dependent outcomes. Under conditions demanding genomic integrity, CHD4 facilitates efficient DNA damage repair, evidenced by its robust pan-cancer correlation with HRD and LOH scores, consistent with its established DNA damage response functions [44–46]. Conversely, in malignancy, this machinery is co-opted to drive oncogenic programs, including stemness maintenance, epithelial-mesenchymal transition, and—as a focal point here—the orchestration of an immunosuppressive microenvironment.

Indeed, our pan-cancer analysis reveals a robust association between elevated CHD4 expression and an immunologically “cold” tumor microenvironment, characterized by features such as diminished CD8 + T cell infiltration, downregulated antigen presentation machinery (e.g., MHC-I), and upregulation of multiple inhibitory checkpoint molecules, including PD-L1, CTLA-4, and LAG-3. These correlative findings are strongly supported by a growing body of functional evidence. Mechanistically, CHD4 appears to orchestrate immune evasion through both cell-autonomous and non-cell-autonomous pathways. In a cell-autonomous manner, CHD4, as part of the NuRD complex, can directly repress the transcription of genes critical for anti-tumor immunity. For instance, in hepatocellular carcinoma, the expression of the CHD4/NuRD complex is negatively correlated with CD8 T cell infiltration, suggesting its potential influence on the tumor immune microenvironment through regulation of complement gene expression [43]. Additionally, CHD4 has been found to modulate early B cell development and V(D)J recombination, which are critical for adaptive immune responses [47]. Concurrently, CHD4 can promote the expression of PD-L1 on tumor cells, possibly through modulating the activity of transcription factors or via interaction with RNA modification pathways (m⁶A), thereby directly engaging immune checkpoint mechanisms [9, 48]. Perhaps more intriguing are the non-cell-autonomous roles of CHD4 within the tumor immune landscape. Crucially, CHD4 is not only expressed in cancer cells but is also functionally vital within immune cell compartments. A demonstrated that CHD4 is a critical regulator of Foxp3 + regulatory T cell (Treg) stability and function. Conditional knockout of CHD4 in murine Tregs leads to fatal systemic autoimmunity and profoundly alters their transcriptional signature [49]. Furthermore, specific oncogenic mutations in CHD4 (e.g., the R975H hotspot mutation found in endometrial cancers) have been implicated in reprogramming tumor-associated macrophages (TAMs) towards an M2-like, pro-tumorigenic, and immunosuppressive phenotype [50]. These findings collectively suggest that high CHD4 activity, whether in malignant or stromal cells, fosters a multi-faceted ecosystem that is hostile to effective anti-tumor immunity. Therefore, while our bioinformatic data establish a significant correlation, the emerging mechanistic literature provides a plausible causal framework. CHD4 likely acts as a central epigenetic enforcer of immune suppression by simultaneously dampening immunogenicity in cancer cells and fortifying the suppressive capacity of key immune cell populations within the TME. Future studies employing cell-type-specific knockout models and spatial transcriptomics will be essential to fully disentangle these complex, compartmentalized contributions of CHD4 to the immune-evasive tumor milieu.

The most translatable insight is the identification of CHD4 as a biomarker for a distinctive “dual-sensitization” phenotype, where CHD4-high tumors show heightened sensitivity to both HDAC inhibitors and conventional DNA-damaging chemotherapeutics. This positions CHD4 as a compelling yet complex therapeutic node. Mechanistically, the vulnerability to HDAC inhibitors likely stems from the role of CHD4 in establishing a state of global histone hypoacetylation via HDAC1/2 [51, 52]. This creates a dependency on epigenetic homeostasis—a form of non-oncogene addiction—thereby sensitizing cells to HDAC inhibition [53]. Pharmacological targeting of HDAC1/2 alters the CHD4-NuRD complex’s exact equilibrium, potentially causing a synthetic lethal effect in CHD4-high malignancies via chromatin disorder and DNA damage accumulation [54, 55]. Critically, the broad chemosensitivity can be explained by the CHD4-associated state of epigenomic instability.Hou et al. has established that CHD4 is an essential regulator of homologous recombination repair, and its deficiency leads to increased genomic instability and DNA damage [56]. Pan et al. further demonstrates that loss of CHD4 function sensitizes cancer cells to PARP inhibitors, providing a direct precedent for the synthetic lethal interaction we infer [57]. While this compromised DNA repair fidelity may fuel tumor progression, it simultaneously opens a unique therapeutic window, rendering CHD4-high tumors more vulnerable to genotoxic agents. This situation establishes a distinct therapeutic vulnerability, indicating that CHD4-high cancers are particularly receptive to combination treatments that combine HDAC inhibitors with conventional chemotherapy. This finding suggests a novel therapeutic strategy: leveraging CHD4 expression as a biomarker to identify patients who may derive exceptional benefit from the combination of epigenetic priming with HDAC inhibitors and conventional DNA-damaging chemotherapy.

However, translating this potential requires addressing several key limitations of the current study. First, while our functional validation in osteosarcoma—a representative model of genomically unstable sarcoma—provides crucial mechanistic proof of concept, the full pan-cancer relevance of CHD4 will be strengthened by extending these experiments to carcinomas where immune evasion is a dominant phenotype (e.g., lung or colorectal cancers), utilizing patient-derived organoids or in vivo models to capture the full complexity of the tumor microenvironment. Second, the prognostic duality and context-dependent therapeutic effects of CHD4 highlight its biological complexity. These “negative” or paradoxical associations are not shortcomings but rather critical clues demanding further investigation into the molecular determinants (e.g., co-mutations, tissue of origin) that switch CHD4’s functional output. Finally, the definitive establishment of causality between CHD4 expression, immune evasion, and therapy response awaits targeted in vivo intervention studies. Future research should prioritize the development of specific CHD4 inhibitors and validate its predictive value in prospective clinical cohorts to translate these findings into precision oncology strategies.

Conclusion

In conclusion, this study delineates the oncogenic landscape of CHD4 across cancers, establishing its association with genomic instability, immune suppression, and a unique therapeutic vulnerability profile. By integrating pan-cancer mapping with mechanistic literature, we refine the understanding of CHD4 from a context-dependent factor to a central epigenetic indicator whose expression signature can inform risk stratification and guide the rational selection of epigenetic and combination therapies. Future efforts should focus on developing specific CHD4-targeting agents and validating these predictive concepts in prospective clinical cohorts to realize the promise of epigenetics-guided precision oncology.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: OY, HG; Methodology: FG, OY, TY, OY, HG; Validation: FG, OY, TY, FK, TF, ZW; Investigation: FG, HG, CY; Data Curation: FK, TF, YS; Visualization: FG, OY, TY, ZZ; Writing-Original Draft Preparation: FG, OY, TY; Writing-Review & Editing: FG, OY, TY, FK, TF, HG, CY; Project Administration: OY, HG; Funding Acquisition: OY, HG; Supervision: OY, HG. All authors contributed to the article and approved the submitted version. All authors have read and approved the submitted manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (82172682 and 82373221), Natural Science Foundation of Chongqing (CSTB2023NSCQ-MSX0472), and Science and Technology Program Project of Enshi Tujia and Miao Autonomous Prefecture, Hubei Province (ESQH20240043 and ESQH20240044).

Data availability

All data generated or analysed during this study are included in this article and its supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.