SMC4/SMAD3/NF-κB axis drives cervical cancer progression and radioresistance via DNA damage repair and immune modulation

Abstract

Objective

Cervical cancer remains a significant global health burden, with the molecular determinants of its progression and therapeutic resistance not fully elucidated. This study aimed to identify DNA damage-related genes with prognostic and functional significance.

Methods

Four cervical cancer GEO datasets were integrated and batch-corrected. Differential expression analysis and WGCNA were performed. Machine learning algorithms (LASSO, SVM, and Random Forest) were used to refine key genes. Functional roles of the pivotal gene SMC4 were investigated using in vitro experiments, including proliferation, colony formation, 3D spheroid assays, phalloidin staining, cell-cycle analysis, immunofluorescence, and analysis of DNA damage and immune-associated pathways under ionizing radiation.

Results

Integration of transcriptomic data revealed 16 candidate genes. Machine learning convergence identified five core genes (CCNB2, CDKN2A, CHEK1, TYMS, and SMC4), all of which were significantly upregulated in tumors. Among these, SMC4 was uniquely associated with poor overall and disease-specific survival. Functional assays demonstrated that SMC4 knockdown under ionizing radiation significantly inhibited cervical cancer cell proliferation, colony formation, invasion, and reduced S-phase cells, and impaired DNA damage repair. Mechanistically, SMC4 was found to upregulate SMAD3, activate NF-κB signaling, and promote PD-L1 expression. Single-cell analysis confirmed SMC4‘s predominant expression in epithelial cells and its association with an altered tumor immune context.

Conclusion

Our study identifies SMC4 as a regulator that promotes radioresistance and potentially modulates the tumor immune microenvironment in cervical cancer, likely by coordinating DNA damage repair and activating the SMAD3-NF-κB pathway. These findings suggest that SMC4 could be a potential therapeutic target for radiosensitization and a candidate biomarker for patient stratification.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-026-07979-5.

Introduction

Cervical cancer remains a pressing global health concern, ranking as the fourth most common malignancy among women and representing a leading cause of cancer-related mortality in low- and middle-income regions [1, 2]. Although the widespread implementation of human papillomavirus (HPV) vaccination and screening programs has markedly reduced disease incidence in developed countries, the overall clinical outcomes for patients with advanced or recurrent cervical cancer remain unsatisfactory [3]. The pathogenesis of cervical cancer is multifactorial, involving persistent HPV infection, accumulation of genetic and epigenetic alterations, and disruption of key regulatory pathways governing cell cycle progression, DNA replication, and apoptosis [4]. Beyond viral oncogenesis, increasing evidence indicates that genomic instability and impaired DNA damage repair (DDR) are central hallmarks of cervical tumorigenesis [5]. Defective DDR mechanisms not only drive malignant transformation but also confer resistance to radiotherapy and chemotherapy, thereby worsening patient prognosis [6, 7]. Emerging studies further suggest that DDR dysregulation can influence anti-tumor immunity by shaping tumor-immune interactions and immune checkpoint signaling [8, 9]. However, the extent and molecular basis of DDR-immune crosstalk in cervical cancer remain poorly defined. Elucidating the molecular determinants that link DDR dysregulation to cervical cancer progression and immune modulation may therefore provide new opportunities for precise molecular diagnosis and targeted therapy.

Despite substantial progress in molecular oncology, the regulatory networks connecting DDR dysfunction with the initiation and evolution of cervical cancer remain incompletely understood. In particular, the interplay between DDR-related molecular programs and the tumor immune microenvironment, and its joint impact on radioresistance and disease progression, has not been systematically investigated. Conventional transcriptomic studies often rely on single-cohort analyses, which are susceptible to batch effects and inter-sample variability, limiting the reproducibility and generalizability of their findings [10]. Furthermore, most prior investigations have focused on individual genes or pathways, overlooking the broader systems-level interactions that orchestrate cervical carcinogenesis and immune regulation. With the advent of high-throughput sequencing technologies and integrative computational approaches, it is now feasible to uncover co-expression networks that reflect the coordinated biological behavior of gene sets associated with tumor phenotypes [11–13]. In particular, weighted gene co-expression network analysis (WGCNA) enables the identification of functionally relevant gene modules [14], while machine learning algorithms such as least absolute shrinkage and selection operator (LASSO) regression, support vector machine (SVM), and random forest (RF) models facilitate robust screening of hub genes with diagnostic and prognostic potential. Leveraging these analytical frameworks offers a promising strategy to systematically explore the interplay between DDR-related molecular networks, immune-associated pathways, and cervical cancer pathobiology.

In this context, the present study aimed to identify and validate DDR-related key genes in cervical cancer through a comprehensive multi-dataset bioinformatics framework. Four independent GEO datasets were integrated and batch-corrected to minimize technical variation. Differentially expressed genes between cervical cancer and normal tissues were identified, followed by WGCNA to determine co-expression modules most closely associated with tumor progression. By intersecting differential expression results, WGCNA modules, and DDR-related gene sets, we identified 16 overlapping genes. Subsequent integrative screening using LASSO, SVM, and RF algorithms revealed five core genes—CCNB2, CDKN2A, CHEK1, TYMS, and SMC4—with strong diagnostic value. Among them, SMC4 emerged as a pivotal gene potentially associated with adverse prognosis and enhanced DDR signaling. Importantly, our integrative analyses further suggest that SMC4 may function at the intersection of DDR regulation and tumor immune modulation. Multidimensional validation, including pan-cancer transcriptomic analysis, immune infiltration assessment, single-cell sequencing, and functional experiments, suggested that SMC4 may promote tumor proliferation, radioresistance, and features of an immunosuppressive tumor microenvironment, although further mechanistic studies are warranted. Collectively, our findings delineate a novel DDR-related molecular signature in cervical cancer and highlight SMC4 as a potential biomarker and therapeutic target with clinical translational relevance.

Methods

Survival and pan-cancer analyses

Prognostic evaluation was conducted in The Cancer Genome Atlas (TCGA)-CESC cohort using Kaplan-Meier survival analysis to assess overall survival (OS) and disease-specific survival (DSS). The log-rank test determined the statistical significance of survival differences. Univariable and multivariable Cox regression analyses were further performed to evaluate whether SMC4 expression was an independent prognostic factor after adjusting for key clinical covariates (age, tumor grade, chemotherapy, and clinical stage). Only variables with a p-value < 0.1 in the univariable analysis were included in the multivariable Cox regression analysis. To examine the broader oncogenic relevance, a pan-cancer analysis across multiple TCGA tumor types was performed, assessing expression patterns of the key genes between tumor and adjacent normal tissues.

Functional annotation and immune infiltration analysis of SMC4

Given its significant prognostic value, SMC4 was selected for in-depth functional characterization. DEGs associated with SMC4 expression were subjected to GO and KEGG enrichment analyses, revealing strong associations with homologous recombination, double-strand break repair, and DNA damage response pathways. Immune microenvironmental characteristics were evaluated using eight established algorithms—ssGSEA [15, 16], quanTIseq [17], TIMER [18], EPIC [19], ESTIMATE [20], MCPcounter [21], CIBERSORT [22], and xCell [23]—to comprehensively delineate immune infiltration patterns.

Cell culture, gene silencing, overexpression, and irradiation

Human cervical cancer cell lines (SiHa and HeLa) were cultured under standard conditions (37 °C, 5% CO₂) in appropriate media supplemented with 10% fetal bovine serum. Small interfering RNAs (siRNAs) targeting SMC4 were transfected using RNATransMate (Sangon) according to the manufacturer’s protocol, and knockdown efficiency was validated by quantitative real-time polymerase chain reaction (qRT-PCR). All siRNA sequences are listed in Supplementary Table S1, and shRNA sequences used for stable knockdown are provided in Supplementary Table S2. At 24 hours post-transfection, cells were exposed to ionizing radiation to evaluate radiosensitivity and downstream functional responses. Irradiation was performed using a Varian TrueBeam linear accelerator delivering 6 MeV X-ray beams at a dose rate of 6 Gy/min. Additionally, for SMC4 overexpression, the full-length coding sequence of SMC4 was cloned into the pMSCV lentiviral vector.

qRT-PCR

Total RNA was isolated from cultured cells using the Eastep™ Total RNA Super Extraction Kit (LS1040, PROMEGA, USA). Complementary DNA (cDNA) was subsequently synthesized from the extracted RNA with the Eastep RT Master Mix (LS2052, PROMEGA, USA). Quantitative real-time PCR (qRT-PCR) was conducted on an ABI 7500 Real-Time PCR System (ABI, USA) employing the Eastep QPCR Master Mix (2×) (LS2062, PROMEGA, USA), following the manufacturer’s protocol. β-actin was used as the internal reference gene for normalization of expression levels. All primer sequences were listed in Supplementary Table S3. Relative transcript abundance between experimental and control samples was calculated using the 2^-ΔΔCt method.

Cell proliferation assays

Cell viability and proliferative capacity under ionizing radiation were evaluated using Cell Counting Kit-8 (CCK-8; ABK0011C-5 ml, ABKBio, China) and colony formation assays, respectively. For the CCK-8 assay, 5000 cells per well were seeded in 96-well plates and treated as indicated, followed by exposure to 6 Gy irradiation; subsequently, absorbance at 450 nm was measured for quantitative analysis. For the colony formation assay, 1000 cells per well were seeded in 6-well plates and irradiated with 6 Gy, and then cultured for 10–14 days. Colonies were fixed, stained with crystal violet, and counted manually. All experiments were performed in triplicate to ensure reproducibility.

Three-dimensional (3D) tumor spheroid formation assay

The three-dimensional tumor spheroid formation assay was performed using cells maintained in the logarithmic growth phase and exhibiting optimal viability. Cells were washed twice with phosphate-buffered saline, enzymatically dissociated with trypsin, centrifuged, and resuspended in complete medium after removal of the supernatant. Cell numbers were determined, and the suspension was adjusted to a final concentration of 1 × 10^5 cells/mL. Subsequently, 50 μL of the cell suspension was dispensed into each well of an ultra-low attachment 96-well round-bottom plate (Corning, NY, USA) using a multichannel pipette, yielding a seeding density of 5000 cells per well. The plates were centrifuged at 500 × g for 3 min to facilitate initial cell aggregation and then placed in a humidified incubator (37 °C, 5% CO₂, 95% relative humidity). Cells were cultured for 48–72 h until compact and morphologically uniform spheroids were established. Once stable spheroid structures had formed, the spheroids were exposed to a single dose of 6 Gy ionizing radiation to evaluate their radiobiological response under 3D culture conditions. At designated time points following irradiation, spheroid morphology and growth dynamics were documented using an inverted optical microscope.

5-ethynyl-2′-deoxyuridine (EdU) incorporation assay

Cell proliferation was evaluated using a 5-ethynyl-2′-deoxyuridine (EdU) incorporation assay. Cells were seeded into 6-well plates at an appropriate density and allowed to adhere and recover overnight under standard culture conditions. A 2× EdU working solution (20 μM) was freshly prepared by diluting a 10 mM EdU stock solution (CX003, EpiZyme, China) at a ratio of 1:500 in pre-warmed complete culture medium. The EdU working solution was then added to each well at a 1:1 ratio with the existing medium, yielding a final EdU concentration of 10 μM. Cells were incubated with EdU for 2 h at 37 °C, with the labeling duration adjusted according to the proliferation characteristics of the cells, corresponding to approximately 10% of the cell cycle. Following EdU incorporation, cells were fixed with 1 mL of fixation buffer per well for 15 min at room temperature. After fixation, cells were washed three times with wash buffer (3–5 min per wash) and subsequently permeabilized with 1 mL of permeabilization buffer for 10–15 min at room temperature. After one to two additional washes, the Click chemistry reaction was performed strictly according to the manufacturer’s instructions. Briefly, the Click Additive Solution was freshly dissolved in 1.5 mL of deionized water, and 0.5 mL of the prepared Click reaction mixture was added to each well. The plate was then incubated in the dark at room temperature for 30 min. Upon completion of the Click reaction, cells were washed three times with wash buffer and counterstained with DAPI (C0065, Solarbio, China) for 10 min at room temperature in the dark. After three additioCnal washes, cells were subjected to fluorescence imaging. EdU incorporation was detected using a 555 Azide dye (excitation/emission: 555/565 nm), while cell nuclei were visualized by DAPI staining (excitation/emission: 360/400 nm).

Cell cycle analysis

The distribution of cells across different phases of the cell cycle was determined using a cell cycle analysis kit (CA1510-20T, Solarbio, China). Cells were first fixed overnight at 4 °C in 70% ice-cold ethanol. On the following day, fixed cells were incubated with RNase A and propidium iodide (PI) for 30 min to allow DNA staining. Cell cycle profiles were subsequently analyzed using a flow cytometer (B49009AD, Beckman, USA) and processed with CytExpert software. PI fluorescence was detected at an excitation wavelength of 488 nm.

Phalloidin-TRITC cytoskeletal staining

Phalloidin-TRITC cytoskeletal staining (CA1620-300T, Solarbio, China) was conducted to assess radiation-induced alterations in cytoskeletal F-actin organization. Cells were seeded on glass coverslips and cultured for over 24 h until reaching approximately 50–60% confluence. Once adhesion was stabilized, cells were irradiated with a single 6 Gy dose of ionizing radiation and subsequently subjected to cytoskeletal staining procedures. Briefly, cultures were rinsed twice with pre-warmed PBS (pH 7.4), fixed in 4% paraformaldehyde for 10–30 min at room temperature, and washed with PBS 2–3 times (10 min each). Permeabilization was performed using 0.5% Triton X-100 for 5 min, followed by an additional 2–3 PBS washes of equal duration. A TRITC-conjugated phalloidin working solution (100 nM, freshly diluted in PBS) was applied to fully cover the cell monolayer, and samples were incubated at room temperature for 30 min in the dark. After staining, coverslips were washed three times with PBS (5 min each) and mounted using antifade mounting medium supplemented with DAPI for nuclear counterstaining. Fluorescence signals were visualized using a fluorescence or confocal microscope equipped with TRITC (Ex/Em = 540/570 nm) and DAPI (Ex/Em = 364/454 nm) filters.

Western blot analysis

Proteins were extracted from cells using RIPA lysis buffer (PC101, EpiZyme, China) supplemented with 1% protease inhibitor cocktail (GRF101, EpiZyme, China) and 1% phosphatase inhibitor cocktail (GRF102, EpiZyme, China). The protein concentrations were determined using a BCA assay kit (ZJ101, EpiZyme, China), and samples were normalized based on their protein content. Denatured protein samples were separated by 4–20% SDS-polyacrylamide gel electrophoresis (SDS-PAGE; ACE, China) and subsequently transferred onto polyvinylidene fluoride membranes (ISEQ00010, Millipore, Germany). Membranes were blocked with 5% skim milk for 1 h at room temperature and then incubated with primary antibodies at 4 °C overnight. Following washes, membranes were incubated with the corresponding HRP-conjugated secondary antibodies for 1 h at room temperature. Primary antibodies employed in this study included Phospho-IKKα/β (CST, 2697, dilution 1:1000), Phospho-IkBα (CST, 2859, dilution 1:1000), Phospho-NF-κB p65 (CST, 3033, dilution 1:1000), SMAD3 (Proteintech, 66,516–1-Ig, dilution 1:2000), SMC4 (Proteintech, 24,758–1-AP, dilution 1:5000), γH2AX (Abcam, ab81299, dilution 1:7500), PD-L1 (Proteintech, 66,248–1-Ig, dilution 1:6000), and the internal reference protein β-actin (ABclonal, AC050, dilution 1:10000). Corresponding horseradish peroxidase-conjugated secondary antibodies (ABclonal, AS014 and AS003, dilution 1:5000) were applied for protein detection. Protein bands were visualized using enhanced chemiluminescence (ECL; MA0186, Meilunbio, China) and imaged with a Bio-Rad imaging system (1708195, USA). Quantitative analysis of band intensities was performed using ImageJ software.

Immunofluorescence assay

Cells in logarithmic growth were seeded onto confocal culture dishes and cultured at 37 °C with 5% CO₂. After adherence, cells were exposed to 6 Gy X-ray irradiation and subsequently subjected to immunofluorescence staining to assess NF-κB nuclear translocation. Following irradiation, cells were washed twice with PBS, fixed with 4% paraformaldehyde for 20 min, and rinsed three times with immunofluorescence washing buffer. Permeabilization was performed for 15 min, followed by three additional washes. Cells were then blocked for 30 min and incubated with a primary antibody against NF-κB (NF-κB p65, CST, 8242, dilution 1:600) at 4 °C overnight. The next day, cells were washed, incubated with a fluorophore-conjugated secondary antibody (Cy3 Goat Anti-Rabbit IgG (H+L), ABclonal, AS007, dilution 1:500) for 1 h at room temperature in the dark, and washed again. Nuclei were counterstained with DAPI, and samples were mounted with antifade mounting medium. Images were acquired using a laser confocal microscope, and NF-κB nuclear translocation was quantified using ImageJ.

Statistical analysis

All statistical analyses were conducted using R software (version 4.3.1) and GraphPad Prism (version 10.1.2). Group comparisons were performed using Student’s t-test or the Wilcoxon rank-sum test, as appropriate. Correlation analyses were conducted using Spearman’s method. Statistical significance was set at p-value < 0.05 unless otherwise specified.

Results

Integration and batch correction of cervical cancer datasets

The overall design of the study is illustrated in Fig. 1. To achieve a comprehensive transcriptional landscape, four cervical cancer-related GEO datasets (GSE122697, GSE14404, GSE64217, and GSE7410) were integrated. PCA demonstrated pronounced separation among the datasets before batch correction (Fig. 2A). To mitigate batch effects and ensure comparability, ComBat-based normalization was applied, resulting in a unified dataset suitable for downstream analyses (Fig. 2B).

Fig. 1.

Overview of the study design and workflow. The overall study design is illustrated, encompassing the integration and batch correction of multiple cervical cancer-related datasets, downstream transcriptomic analyses, and experimental validation. The workflow includes single-cell and bulk RNA-seq processing, identification of cell populations, differential gene expression and co-expression network analyses, machine learning-based selection of key regulatory genes, functional enrichment and immune infiltration analyses with a focus on SMC4, in vitro and 3D tumor spheroid experiments under ionizing radiation to assess proliferation, invasion, and cytoskeletal remodeling, and mechanistic investigation of the SMC4/SMAD3/NF-κB axis, culminating in a proposed model of SMC4-mediated tumor progression, immune evasion, and radioresistance. scRNA-seq, single-cell RNA sequencing; 3D, three-dimensional

Fig. 2.

Batch correction and differential gene expression analysis in cervical cancer datasets. (A) PCA of four GEO datasets (GSE122697, GSE14404, GSE64217, and GSE7410) before batch effect correction. (B) PCA after batch correction, demonstrating successful reduction of inter-dataset variability. (C) Heatmap showing DEGs between normal cervical tissues and cervical cancer tissues. (D) Volcano plot representing upregulated and downregulated DEGs. (E-F) Functional enrichment analysis of DEGs, highlighting significant involvement in DNA replication, mitosis, cell division, nuclear division, and cell-cycle regulation. PCA, principal component analysis; GEO, gene expression Omnibus; DEGs, differentially expressed genes

Differential expression analysis and functional enrichment

Differential gene expression analysis between normal cervical tissues and cervical cancer samples revealed a substantial set of DEGs. Heatmaps and volcano plots illustrated the significant transcriptional differences (Fig. 2C–D). Functional enrichment analysis indicated that these differentially expressed genes were predominantly involved in DNA replication, mitotic division, chromosome segregation, nuclear division, and cell-cycle regulation (Fig. 2E–F), underscoring the pivotal roles of cell proliferation and genomic instability in cervical tumorigenesis.

Identification of key modules via WGCNA

WGCNA was conducted to identify modules of co-expressed genes and correlate them with disease phenotypes. The turquoise module exhibited the strongest association with cervical cancer (Fig. 3A–C). Functional annotation suggested enrichment in DNA damage response pathways. The intersection of differentially expressed genes, hub genes from WGCNA, and DNA damage repair-associated genes yielded 16 candidate genes, all upregulated in tumor tissues (Fig. 3D–E). Correlation analysis revealed a tightly interconnected network among these 16 genes (Fig. 3F).

Fig. 3.

Identification of key gene modules via WGCNA and intersection with DNA damage-related genes. (A-C) WGCNA identified gene modules associated with cervical cancer phenotypes, with the MEturquoise module showing the strongest correlation. (D) Venn diagram illustrating 16 intersecting genes derived from DEGs, WGCNA-selected module genes, and DNA damage repair-related genes. (E) Expression levels of the 16 intersecting genes in normal versus cervical cancer tissues, showing higher expression in tumors. (F) Correlation matrix of the 16 intersecting genes, indicating co-expression relationships. WGCNA, weighted gene co-expression network analysis; DEGs, differentially expressed genes

Chromosomal distribution and machine learning-based hub gene selection

Mapping the 16 candidate genes to the genome revealed that all were located on autosomes (Fig. 4A–B). To refine the key regulatory genes, three machine-learning approaches—LASSO regression, SVM, and RF—were employed. LASSO identified seven genes (Fig. 4C–D), SVM yielded fifteen (Fig. 4E–F), and RF highlighted eleven (Fig. 4G–H). The intersection of these results pinpointed five robust key genes: CCNB2, CDKN2A, CHEK1, TYMS, and SMC4 (Fig. 5A), all exhibiting significantly higher expression in cervical cancer tissues relative to normal controls (Fig. 5B).

Fig. 4.

Chromosomal location and machine-learning-based screening of candidate genes. (A-B) Chromosomal distribution of the 16 intersecting genes, demonstrating localization to autosomes. (C-D) LASSO regression identified 7 key genes. (E-F) SVM analysis identified 15 key genes. (G-H) RF analysis identified 11 key genes

Fig. 5.

Expression, diagnostic performance, and predictive modeling of the five hub genes. (A) Venn diagram showing the overlap of candidate genes from three machine-learning algorithms. (B) Comparison of expression levels between normal and cervical cancer tissues for the five hub genes. (C-D) ROC curves demonstrating individual and combined diagnostic accuracy of the hub genes, with a combined AUC of 0.903. (E) Nomogram for predicting cervical cancer risk based on the five hub genes. (F) Calibration curve showing consistency between predicted and observed probabilities. (G) DCA demonstrating clinical applicability of the nomogram. (H-K) Validation of gene expression and ROC performance in independent GEO cohorts (GSE9750 and GSE67522). ROC, receiver operating characteristic; AUC, area under the curve; DCA, decision curve analyses; GEO, gene expression Omnibus

LASSO, least absolute shrinkage and selection operator; SVM, support vector machine. LASSO, least absolute shrinkage and selection operator; SVM, support vector machine; RF, random forest.

Diagnostic and prognostic assessment of key genes

Receiver operating characteristic (ROC) analysis demonstrated strong diagnostic performance for the five selected genes, with individual AUC values exceeding 0.8 and a combined AUC of 0.903 (Fig. 5C–D). A nomogram incorporating these genes displayed excellent calibration and potential clinical utility, as evidenced by calibration and decision curve analyses (Fig. 5E–G). Validation using independent GEO datasets (GSE9750 and GSE67522) confirmed the upregulation and high diagnostic accuracy of these genes (Fig. 5H–K). Survival analysis in the TCGA-CESC cohort revealed that only SMC4 significantly correlated with OS and DSS (Supplementary Fig. S1A-J). We confirmed using the TCGA database that the expression levels of these five genes were significantly higher in cervical cancer tissues compared to normal cervical tissues (Supplementary Fig. S1K). Pan-cancer analysis further demonstrated that SMC4 was overexpressed across multiple tumor types (Supplementary Fig. S1L).

Functional enrichment and immune profiling of SMC4

Functional enrichment of SMC4-associated genes highlighted involvement in homologous recombination, double-strand break repair, DNA damage response, and cell-cycle regulation (Supplementary Fig. S2A-D), confirming its role in maintaining genomic stability. Multi-algorithm immune infiltration analyses, including ssGSEA, Quantiseq, TIMER, EPIC, ESTIMATE, MCP-counter, CIBERSORT, and xCell, revealed that high SMC4 expression was associated with reduced T-cell infiltration and lower immune scores, suggesting a potential link to an immunosuppressive tumor microenvironment rather than establishing a direct causal role (Supplementary Fig. S2E-J and Supplementary Fig. S3A-C).

SMC4 is an independent prognostic factor for overall and disease-specific survival in cervical cancer

In order to determine whether SMC4 serves as an independent prognostic indicator, univariable and multivariable Cox regression analyses were performed for OS and DSS, including SMC4 expression together with key clinical covariates (age, tumor grade, chemotherapy, and clinical stage) as adjustment variables. For OS, univariable analysis showed that age, clinical stage, and SMC4 expression had p-values < 0.1, indicating potential associations with survival. When these factors were included in the multivariable model, only clinical stage [hazard ratio (HR): 4.078; 95% CI: 2.180–7.630; p < 0.001] and SMC4 expression (HR: 1.818; 95% CI: 1.005–3.289; p = 0.048) remained independent prognostic factors (Supplementary Table S4). For DSS, univariable analysis showed that chemotherapy, clinical stage, and SMC4 expression had p-values < 0.1, indicating potential associations with DSS. In the multivariable analysis, clinical stage (HR: 3.822; 95% CI: 1.818–8.036; p < 0.001) and SMC4 expression (HR: 2.174; 95% CI: 1.082–4.368; p = 0.029) were independently associated with DSS, confirming the independent prognostic value of SMC4 in cervical cancer (Supplementary Table S5).

SMC4 knockdown suppresses proliferation and clonogenicity of cervical cancer cells under ionizing radiation

Baseline SMC4 expression was first evaluated in cervical cancer cell lines and a non-tumorigenic cervical epithelial control. Western blot analysis showed that SMC4 protein levels were markedly elevated in both SiHa and HeLa cells compared with Ect1/E6E7 cells (Supplementary Fig. S4A). To elucidate the biological function of SMC4, we constructed SMC4 interference vectors and identified siRNA-2 as the most efficient sequence for knockdown in both SiHa and HeLa cells (Fig. 6A–B). Western blot analysis further confirmed the knockdown efficiency of SMC4 across NC, siRNA-1, siRNA-2, and siRNA-3 groups, with siRNA-2 showing the most pronounced reduction in SMC4 protein levels, consistent with the qPCR results (Fig. 6C). To further characterize the temporal dynamics of SMC4 silencing, Western blot analysis was performed at multiple time points following siRNA transfection (24, 36, 60, 84, and 120 h). SMC4 protein reduction was detectable as early as 24 h post-transfection, reached maximal suppression at intermediate time points, and gradually diminished thereafter, while remaining partially suppressed at 120 h (Supplementary Fig. S4B-C). Under ionizing radiation, CCK-8, colony formation, and EdU assays collectively demonstrated that knockdown of SMC4 using siRNA-2 markedly suppressed the proliferative capacity of cervical cancer cells (Fig. 6D–H). Furthermore, 3D tumor- spheroid formation assays confirmed that SMC4 knockdown substantially impaired both spheroid formation ability and subsequent growth kinetics (Fig. 6I). To further validate the function of SMC4, we also used the siRNA-3 sequence for knockdown. Our CCK-8 results similarly demonstrated that SMC4 knockdown with siRNA-3 also inhibited the proliferative capacity of cervical cancer cells under ionizing radiation conditions (Supplementary Fig. S4D-E). To address concerns regarding the durability of transient siRNA-mediated knockdown in long-term assays, colony formation experiments were performed using cells with stable SMC4 suppression. Western blot analysis (Fig. 7A) confirmed that shSMC4 efficiently reduced SMC4 protein levels in both SiHa and HeLa cells. Furthermore, protein expression was evaluated at two time points following shSMC4 transfection (72 and 168 h), with significant reduction of SMC4 observed at both time points (Supplementary Fig. S4F-G). Consistently, as shown in Supplementary Fig. S4H-I, shSMC4-expressing cells formed significantly fewer colonies compared with the corresponding control cells, further demonstrating that sustained inhibition of SMC4 markedly impaired the clonogenic potential of cervical cancer cells. Our EdU experiment results similarly confirmed that knockdown of SMC4 with siRNA-3 inhibited the proliferative capacity of cervical cancer cells (Supplementary Fig. S4J). Furthermore, 3D tumor spheroid formation assays confirmed that SMC4 knockdown with siRNA-3 significantly impaired spheroid formation ability and subsequent growth kinetics (Supplementary Fig. S4K).

Fig. 6.

Knockdown of SMC4 suppresses the proliferation of cervical cancer cells under ionizing radiation. (A-C) Verification of SMC4 knockdown efficiency in SiHa and HeLa cells using three candidate siRnas; siRNA-2 exhibited the most effective silencing and was selected for subsequent experiments. (D-E) CCK-8 assays revealed that SMC4 silencing significantly reduced cell viability following 6 Gy ionizing radiation (cell viability was measured at 24, 48, and 72 hours post-irradiation). (F-H) Colony formation and EdU incorporation assays further demonstrated that SMC4 knockdown markedly impaired the proliferative capacity of cervical cancer cells under radiation (for colony formation, cells were irradiated, then cultured for 10–14 days before staining and counting; for EdU assay, cells were incubated with EdU at 24 hours post-irradiation before fixation and detection). (I) Three-dimensional tumor spheroid formation assays showed that SMC4 depletion substantially decreased both spheroid formation efficiency and growth kinetics (following siRNA transfection, cells were cultured to allow spheroid formation for 24–48 hours and then were irradiated)

Fig. 7.

SMC4 promotes DNA damage repair and immune evasion under ionizing radiation. (A) Western blot analysis confirmed that shSMC4 efficiently reduced SMC4 protein levels in both SiHa and HeLa cells. (B-C) γH2AX expression was analyzed at multiple time points (2, 4, 8, 12, and 24 h) following irradiation. SMC4 knockdown increased γH2AX levels and delayed their resolution, whereas SMC4 overexpression reduced γH2AX levels, indicating promotion of radioresistance. (D-F) Analysis of public cervical cancer transcriptomic datasets (GSE6791, GSE9750, and GSE151666) showed that SMC4 expression did not differ significantly between HPV-positive and HPV-negative patients, suggesting that SMC4 expression was largely independent of HPV status. (G-H) Western blot analysis confirmed that SMC4 expression levels in wild-type (WT) and vector control (VC) cells were comparable, whereas SMC4-overexpressing cells exhibited elevated SMC4 protein levels within the expected experimental range. (I-J) Mechanistic analysis revealed that SMC4 knockdown inhibited SMAD3 expression, suppressed NF-κB signaling activation, and reduced PD-L1 levels. (K-M) Pan-cancer immune infiltration analysis showed that higher SMAD3 expression correlated with lower immune scores and reduced CD8+ T cell infiltration, suggesting a role for SMC4 in immune evasion. WT, wild-type; VC, vector control

SMC4 regulates invasive behavior and clonogenic growth in cervical cancer cells

Phalloidin staining demonstrated that SMC4 silencing markedly reduced filopodia formation and inhibited cell invasive behavior under ionizing radiation (Supplementary Fig. S5A), suggesting a role for SMC4 in cytoskeletal remodeling and metastatic potential. To assess the specificity of SMC4-mediated effects, a dose-dependent rescue strategy was implemented using an siRNA-resistant SMC4 construct. Under SMC4 knockdown conditions, graded re-expression of SMC4 was achieved using a common CMV promoter with increasing transfection amounts (SMC4-low and SMC4-mid). Western blot analysis confirmed the stepwise restoration of SMC4 protein levels in the siSMC4 + SMC4-low and siSMC4 + SMC4-mid groups compared with siSMC4 alone (Supplementary Fig. S5B). Consistently, clonogenic survival after irradiation was progressively rescued in a manner proportional to SMC4 expression levels, with partial recovery observed in the SMC4-low group and further improvement in the SMC4-mid group (Supplementary Fig. S5C). These results indicate a quantitative relationship between SMC4 protein abundance and post-irradiation clonogenic capacity, supporting a contributory role of SMC4 in radioresistance.

SMC4 may modulate DNA damage response irrespective of cell-cycle distribution and HPV status

The role of SMC4 in DNA damage repair was investigated by analyzing γH2AX expression in NC and shSMC4 cells at multiple time points (2, 4, 8, 12, and 24 h) following irradiation in both SiHa and HeLa cells. γH2AX levels were elevated in shSMC4 cells compared with NC, with both groups peaking at 4 h post-irradiation and gradually decreasing thereafter (Fig. 7B–C), suggesting that SMC4 depletion delayed the resolution of DNA double-strand breaks and implicated SMC4 in the regulation of DNA damage response kinetics. Analysis of public cervical cancer transcriptomic datasets (GSE6791, GSE9750, and GSE151666) revealed that SMC4 expression did not differ significantly between HPV-positive and HPV-negative patients (Fig. 7D–F), suggesting that SMC4 expression was not strongly associated with HPV status. Flow cytometry-based cell-cycle profiling was performed to determine whether the effects of SMC4 on radiosensitivity were linked to cell-cycle alterations. Before irradiation, SMC4 knockdown reduced the proportion of S-phase cells and increased G2/M-phase cells. Following irradiation, however, no significant differences in cell-cycle distribution were observed between control and SMC4-silenced cells (Supplementary Fig. S5D-E), suggesting that the enhanced radiosensitivity associated with SMC4 depletion was largely independent of gross cell-cycle redistribution.

SMC4 may modulate radioresistance and immune evasion via the SMAD3-NF-κB signaling pathway

Western blot analysis confirmed that SMC4 expression levels in both wild-type (WT) and vector control (VC) cells were comparable, whereas cells stably transduced with the lentiviral SMC4 overexpression construct exhibited elevated SMC4 levels relative to VC, remaining within the expected experimental range (Fig. 7G–H). Knockdown of SMC4 under irradiation increased γH2AX levels, indicating enhanced DNA damage, whereas SMC4 overexpression reduced DNA damage, thereby promoting radioresistance. Mechanistic analysis revealed that SMC4 knockdown inhibited SMAD3 expression, suppressed NF-κB signaling activation, and reduced PD-L1 levels (Fig. 7I–J). Notably, under irradiation conditions, SMC4 knockdown decreased NF-κB nuclear translocation (Supplementary Fig. S5F). Co-immunoprecipitation experiments demonstrated a physical interaction between SMC4 and SMAD3 in both SiHa and HeLa cells (Supplementary Fig. S5G-H), indicating that SMC4 may promote tumor progression and radioresistance through the SMAD3-NF-κB axis. Pan-cancer immune infiltration analysis of SMAD3 revealed that higher SMAD3 expression was associated with lower immune scores and reduced CD8+ T cell infiltration in most tumors (Fig. 7K–M). Analysis of two immunotherapy-treated cancer cohorts (GSE207422 and GSE289743) indicated that patients with high SMC4 expression tended to exhibit poorer responses to anti-PD1/PD-L1 therapy, suggesting a possible predictive association that warrants further experimental validation (Supplementary Fig. S5I-J).

SMC4 expression and cellular heterogeneity in cervical cancer revealed by single-cell analysis

Eight single-cell RNA-seq datasets (GSE168652, GSE173231, GSE197461, GSE224327, GSE236738, GSE279998, E-MTAB-11948, and E-MTAB-12305) comprising 58 cervical cancer samples were analyzed. Clustering identified ten major cell populations, including epithelial cells, fibroblasts, macrophages, endothelial cells, B cells, T cells, myeloid cells, mast cells, stromal cells, and VSMCs (Fig. 8A–B). Figure 8Cshowed the proportions of each cell subpopulation in individual samples. Tumor tissues exhibited higher epithelial and macrophage fractions compared to normal tissues (Fig. 8D). We further analyzed the distribution of SMC4 across these cell subpopulations and found that SMC4 was predominantly localized in epithelial cells (Fig. 8E). We also confirmed that SMC4 expression levels were significantly higher in cervical cancer tissues compared to normal cervical tissues (Fig. 8F–G). Additionally, SMC4-positive cells exhibited a markedly higher proportion of epithelial cells and lower proportions of B cells and T cells compared to SMC4-negative cells (Fig. 8H–I). Cell-cycle profiling indicated enrichment of S and G2/M phases in tumor cells (Fig. 8J). Finally, Fig. 8K illustrated the proposed mechanistic model based on the findings of this study.

Fig. 8.

SMC4 expression and cellular heterogeneity in cervical cancer revealed by single-cell RNA-seq analysis. (A-B) UMAP visualization of single-cell transcriptomes from 58 cervical cancer samples across eight datasets, showing ten major cell populations: epithelial cells, fibroblasts, macrophages, endothelial cells, B cells, T cells, myeloid cells, mast cells, stromal cells, and vascular smooth muscle cells. (C) Proportions of each cell subpopulation in individual samples. (D) Comparison of cell-type composition between tumor and normal tissues, revealing higher epithelial and macrophage fractions in tumor tissues. (E) Distribution of SMC4 expression across cell populations, predominantly in epithelial cells. (F-G) SMC4 expression levels were significantly elevated in cervical cancer tissues relative to normal cervical tissues. (H-I) Comparison of cell-type composition between SMC4-positive and SMC4-negative cells, showing enrichment of epithelial cells and a reduction of B and T cells in SMC4-positive cells. (J) Cell-cycle phase analysis indicated enrichment of S and G2/M phases in tumor cells. (K) Proposed mechanistic model summarizing the roles of SMC4 in cervical cancer progression. scRNA-seq, single-cell RNA sequencing; UMAP, uniform manifold approximation and projection

Discussion

The pathogenesis of cervical cancer, while predominantly linked to high-risk HPV infection, involves complex molecular alterations that extend beyond viral oncogene expression. Our study provides a comprehensive multi-omics dissection of these alterations, culminating in the identification of a five-gene diagnostic signature and the functional characterization of SMC4 as a pivotal oncoprotein. The initial integration of multiple microarray datasets, corrected for batch effects, provided a robust foundation for discovering consistently dysregulated pathways. The significant enrichment of differentially expressed genes in DNA replication, mitosis, and cell-cycle regulation is a hallmark of HPV-driven carcinogenesis, where the E6 and E7 oncoproteins directly abrogate cell-cycle checkpoints, inducing replicative stress and genomic instability [5]. We acknowledge that this initial transcriptional landscape is heavily shaped by the pervasive effects of HPV. However, to distinguish central drivers from secondary phenomena, we employed a stringent, multi-layered filtering strategy. Our approach of intersecting these deregulated genes with a co-expression module (MEturquoise) linked to DNA damage repair createed a powerful filter, pinpointing genes that are not merely correlative but potentially central to the core vulnerabilities of cervical cancer cells.

The application of three distinct machine-learning algorithms (LASSO, SVM, RF) to refine the 16 initial candidates represents a rigorous, consensus-driven strategy to minimize false positives and identify the most robust biomarkers. The resultant five-gene panel—CCNB2, CDKN2A, CHEK1, TYMS, and SMC4—encapsulates key cancer hallmarks. As astutely noted, genes such as CCNB2, CHEK1, and TYMS are known to be influenced by E2F activation downstream of HPV E7-mediated Rb inactivation. This overlap underscores the relevance of our signature to cervical cancer biology but also raises the important question of whether these genes, particularly SMC4, represent independent drivers or epiphenomena of HPV infection. To address the potential confounding influence of HPV status, we performed a critical additional analysis. Interrogation of three independent GEO datasets with clear HPV annotation revealed no significant difference in SMC4 expression between HPV-positive and HPV-negative tumors. This finding suggests that SMC4 upregulation is not a mere surrogate for HPV positivity but may be driven by other oncogenic mechanisms common across cervical cancer subtypes. This HPV-status independence strengthens the case for SMC4 having a fundamental role beyond the immediate scope of viral oncogene action. CCNB2 is a core mitotic regulator [24]; CDKN2A, though often inactivated in cancers, can exhibit context-dependent oncogenic functions or reflect stromal expression [25]; CHEK1 is a critical kinase in the DNA damage response [26]; and TYMS is essential for nucleotide biosynthesis, fueling rapid proliferation [27]. Their collective strong diagnostic performance, validated in independent cohorts, and the clinical utility of the resulting nomogram suggest this signature could augment current diagnostic methodologies.

Strikingly, among this potent signature, only SMC4 demonstrated significant prognostic value for both overall and disease-specific survival in the TCGA cohort, and it remained an independent prognostic factor after adjusting for key clinical covariates including clinical stage. SMC4, a core component of the condensin complex, is indispensable for chromosome condensation and faithful segregation during mitosis [28]. Its overexpression across multiple cancers, as highlighted by our pan-cancer analysis, suggests it is a fundamental mitotic regulator whose dysregulation is poorly tolerated by cells, often leading to aneuploidy and further genomic instability [29]. Our functional enrichment analysis of SMC4-correlated genes provides a mechanistic link, firmly embedding it within networks of homologous recombination and double-strand break repair. Moreover, our in vitro experiments demonstrated that SMC4 promotes DNA damage repair under ionizing radiation, facilitating radioresistance. Time-course analysis of γH2AX foci showed that SMC4 knockdown delayed the resolution of DNA double-strand breaks, indicating impaired repair efficiency. While this confirms SMC4’s role in the DDR, the specific pathways (HR, NHEJ) or checkpoint mechanisms involved remain to be defined. Future studies using pathway-specific reporters and checkpoint assays are needed to clarify its precise mechanistic role. Notably, this radioresistance phenotype was not associated with marked cell-cycle redistribution after irradiation, suggesting that cell-cycle effects alone did not account for the observed response. This effect is mediated through the upregulation of SMAD3, activation of the NF-κB signaling pathway, and induction of PD-L1, forming a potential SMC4–SMAD3-NF-κB axis that coordinates DNA repair and contributes to features of an immunosuppressive microenvironment. While co-immunoprecipitation experiments confirmed a direct interaction between SMC4 and SMAD3, future studies employing ChIP, promoter-reporter, and targeted rescue/inhibition assays are required to definitively establish the directionality and causality of this signaling pathway. This positions SMC4 not just as a mitotic factor, but as a potential key player in managing the replication stress and DNA damage that are quintessential features of HPV-transformed cells, possibly creating a synthetic lethal vulnerability [30].

Beyond cell-autonomous effects, our study points to a potential dimension of SMC4‘s role in shaping the tumor immune microenvironment. The consistent inverse correlation between high SMC4 expression and both CD4+ T cell infiltration and overall immune scores, corroborated by bulk and single-cell analyses, suggests that SMC4-high tumors may be associated with an immunologically “cold” phenotype. This could be mediated through the secretion of specific cytokines or the recruitment of immunosuppressive cells, a hypothesis supported by the observed increase in macrophages in tumor samples. The single-cell data further sharpen this picture, demonstrating that SMC4 expression is predominantly epithelial and that SMC4-positive cellular neighborhoods are characterized by a depletion of lymphocytes. This correlation warrants investigation into SMC4 as a potential biomarker for immune exclusion, with possible implications for predicting responses to immunotherapy [31]. The link between the SMC4/SMAD3/NF-κB axis and PD-L1 upregulation provides a potential mechanistic connection that may contribute to this immunosuppressive microenvironment.

The most direct evidence of SMC4‘s therapeutic relevance comes from our in vitro functional assays. SMC4 knockdown markedly sensitized cervical cancer cells to ionizing radiation, as evidenced by reduced proliferation, colony formation, 3D spheroid growth, and increased DNA damage (γH2AX accumulation). Conversely, SMC4 overexpression conferred radioresistance. Our analysis of DNA damage response kinetics provided evidence of impaired double-strand break repair upon SMC4 knockdown, a key mechanism underlying radiosensitivity. Radiotherapy is a cornerstone for treating locally advanced cervical cancer, and radioresistance remains a major cause of treatment failure [32]. Our data suggest that SMC4 overexpression might be a mechanism by which tumor cells mitigate radiation-induced DNA damage, and its inhibition could be a potent radiosensitization strategy. While clonogenic survival assays remain a gold standard for quantifying radiosensitivity, our integrated approach—encompassing long-term growth inhibition and direct DNA repair assessment—provides functional validation of SMC4‘s role. In future studies, formal clonogenic survival assays could be employed to obtain precise radiobiological parameters under SMC4 modulation, providing further quantitative validation of its impact on radiosensitivity. This positions SMC4 not only as a biomarker but also as a promising therapeutic target to enhance the efficacy of standard treatment. Notably, the dual role of SMC4 in promoting radioresistance and its potential association with an immunologically distinct tumor microenvironment nominates it as a unique target whose inhibition might yield a dual therapeutic benefit: sensitizing tumors to radiation while possibly modulating immune-related features of the tumor milieu. Future development of SMC4-specific inhibitors would be required to test this hypothesis. However, careful consideration must be given to potential on-target toxicities given SMC4‘s essential role in chromosome segregation in normal dividing cells. Strategies such as tumor-selective delivery, intermittent dosing, or combination with immunotherapy may help mitigate such risks and unlock the therapeutic potential of targeting this multifaceted regulator.

While our study provides compelling evidence, several limitations warrant consideration. First, although batch correction and multiple datasets were used, heterogeneity in patient cohorts and sample preparation may influence the observed expression patterns. Second, although the present analyses suggest that the observed role of SMC4 is not strictly dependent on HPV status, further studies incorporating more detailed virological information—such as HPV genotype, viral load, and integration status—in larger, well-annotated cohorts would be valuable to more comprehensively explore potential interactions. Third, while our post-irradiation cell-cycle analysis and SMC4-rescue experiments support its specific role beyond general essentiality, future studies employing cell-cycle-synchronized models or real-time phase-specific reporters could further refine the understanding of SMC4‘s function across the cell cycle. Fourth, the clinical predictive value of the five-gene signature and the SMC4-immunosuppression link, which are currently supported primarily by computational analyses, require validation in large, prospectively collected clinical cohorts with associated treatment and outcome data. Specifically, the observed associations between SMC4 and immune contexture remain correlative; direct experimental validation using immune co-culture or in vivo models is needed to establish causality. Fifth, the precise molecular mechanism by which SMC4 influences DNA damage repair and immune cell infiltration remains elusive. Finally, our in vitro models, while informative, cannot fully recapitulate the complexity of the tumor microenvironment. The use of patient-derived organoids or in vivo models would be necessary to fully validate the functional and therapeutic insights.

Conclusion

In conclusion, our integrative analysis moves beyond a simple gene list to provide a layered understanding of cervical cancer biology. We define a high-performance diagnostic signature and, within it, identify SMC4 as a regulator operating at the nexus of cell-cycle progression, genome integrity, and tumor immunology. Its ability to confer radioresistance in vitro and its association with poor prognosis and an immune-desert TME are mediated, at least in part, through the SMC4/SMAD3/NF-κB signaling axis, elevating it from a candidate gene to a compelling target for future therapeutic strategies aimed at overcoming treatment resistance in cervical cancer.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

HPV

Human papillomavirus

DDR

DNA damage repair

WGCNA

Weighted gene co-expression network analysis

LASSO

Least absolute shrinkage and selection operator

SVM

Support vector machine

GEO

Gene Expression Omnibus

PCA

Principal component analysis

DEGs

Differentially expressed genes

GO

Gene Ontology

KEGG

Kyoto Encyclopedia of Genes and Genomes

SVM-RFE

Support vector machine-recursive feature elimination

RF

Random forest

ROC

Receiver operating characteristic

AUC

Area under the curve

DCA

Decision curve analyses

TCGA

The Cancer Genome Atlas

OS

Overall survival

DSS

Disease-specific survival

HR

Hazard ratio

3D

Three-dimensional

WT

Wild-type

VC

Vector control

scRNA-seq

Single-cell RNA sequencing

UMAP

Uniform manifold approximation and projection

VSMCs

Vascular smooth muscle cells

siRNAs

Small interfering RNAs

qRT-PCR

Quantitative real time-polymerase chain reaction

CCK8

Cell Counting Kit-8

ROC

Receiver operating characteristic

EdU

5-Ethynyl-2′-deoxyuridine

Author contributions

HW, YY, WW, and QX designed this study. HW, YY, and WW contributed to the data collection. HW and YY analyzed the data. HW and QX supervised the study. HW, YY, and QX wrote the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the Joint Funds for the Natural Science Foundation of Fujian Province (2024J011099), Fujian Cancer Hospital High-Level Talent Training Program (2023YNG09), innovation of science and Technology, Fujian province (2023Y9449), Fujian Province Medical Innovation Research Grant (2024CXA031), Clinical Research Grant of Fujian Cancer Hospital (202412004), Natural Science Foundation of Fujian Province (2025J011192), and Clinical Research Center for Precision Treatment of Gynecological Malignancies of Fujian Province (2022Y2015).

Data availability

The datasets supporting the conclusions of this study are available from the corresponding author upon reasonable request.

Declarations

Ethical approval and consent to participate

Not applicable.

Consent for publication

All authors have read and approved the manuscript for publication.

Competing interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.