Elucidating the dynamic tumor microenvironment through deep transcriptomic analysis and therapeutic implication of MRE11 expression patterns in hepatocellular carcinoma.

Introduction
The product of the MRE11 gene is a critical component of the MRE11-RAD50-NBS1 (MRN) complex, primarily responsible for recognizing and repairing DNA double-strand breaks [1]. Recently, MRE11 has attracted significant attention in various cancers [2], with studies indicating its role in genomic stability and tumor development in breast cancer [3], prostate cancer [4], and ovarian cancer [5]. However, the specific role of MRE11 in HCC remains inadequately explored. Existing literature indicates that MRE11 is highly expressed in HCC tissues and may be associated with tumor progression and patient prognosis. However, the precise molecular mechanisms by which it regulates the TME in liver cancer are still unclear [6]. Further investigation into the role of MRE11 in HCC may reveal its potential as a novel therapeutic target.
Beyond its role in DNA damage repair (DDR), MRE11 also modulates the immune microenvironment in cancer [7, 8], particularly in HCC, where it may influence tumor progression by regulating immune cell infiltration and activity [9]. The relationship between MRE11 and the cGAS-STING pathway has garnered considerable interest. The cGAS-STING signaling axis consists of cyclic GMP–AMP synthase (cGAS), which generates the second messenger cyclic GMP–AMP (cGAMP), and the cyclic GMP–AMP receptor stimulator of interferon genes (STING) [10]. Recent studies suggest that MRE11 can promote tumorigenesis by releasing the inhibitory sequestration of cGAS by nucleosomes, thereby activating the cGAS-STING signaling pathway and innate immune response. This regulatory mechanism underscores the pivotal role of MRE11 in linking tumor biology with innate immune surveillance [10, 11]. In HCC, MRE11 may reshape the tumor immune microenvironment by modulating the cGAS-STING pathway, potentially impacting immunotherapy outcomes [12, 13]. Consequently, MRE11 plays a crucial role not only in DNA repair and maintaining genomic stability but also in regulating immune evasion and immune responses in HCC.
Given the limited research on MRE11 in HCC, this study combines multi-layered bioinformatics analyses (including scRNA-seq, Bulk RNA sequencing, Spatial transcriptome analysis, and Mendelian randomization) with experimental validation to comprehensively investigate MRE11’s function and its dynamic role within the HCC tumor microenvironment. Additionally, we developed an HCC prognosis prediction model based on large-scale HCC datasets, employing a combination of 101 machine-learning algorithms. This model successfully identified robust potential prognostic indicators for predicting patient outcomes. Through detailed exploration of molecular mechanisms and the development of clinical prediction tools, MRE11 is poised to become a potential target for HCC diagnosis and personalized treatment.

Materials and methods

Patient sample collection and ethical approval
This study was approved by the Ethics Committee of the General Hospital of the People’s Liberation Army of China (Approval No. S2016-098-02). Tumor and adjacent non-tumor liver tissue samples were collected from 92 HCC patients, with 45 cases from the Fifth Medical Center and 47 cases from the First Medical Center, respectively. Patient demographics and clinical characteristics are detailed in Table 1.

Data for bulk RNA-seq analysis were obtained from The Cancer Genome Atlas (TCGA), Genotype-Tissue Expression (GTEx), International Cancer Genome Consortium (ICGC), and Gene Expression Omnibus (GEO) databases. Machine learning data included the raw mRNA expression profiles from the GEO dataset GSE14520 [14], TCGA dataset, and ICGC LIRI-JP cohort. The TCGA-LIHC dataset (n = 371) was used as the training set for model construction, while GSE14520 (n = 221) and ICGC-LIRI-JP (n = 232) were used as external validation cohorts. For single-cell analysis, we utilized three external scRNA-seq datasets: GSE149614 [15], GSE151530 [16], and GSE125449 [17]. Additionally, spatial RNA transcript data were obtained from a previous publication [18, 19].

Mendelian randomization analysis
We performed Summary-data-based Mendelian Randomization (SMR) analysis using the SMR R package and Two-sample MR analysis using the TwoSampleMR R package. The exposure factor was the MRE11 gene (ID: prot-a-1939), and the outcome factors included HCC (ID: bbj-a-158) and p53/DNA damage-related proteins (ID: prot-a-2238). All data were sourced from the Integrative Epidemiology Unit (IEU) database. To ensure analytical precision, we filtered exposure data with a p-value < 1 × 10−5and excluded linkage disequilibrium between instrumental variables (IVs) using an r2 < 0.001 threshold. We further assessed potential confounders or risk factors for HCC to ensure the robustness of our results.

Diagnostic and prognostic analysis of MRE11
Genes with potential causal association with HCC were screened out by SMR analysis, and MRE11A (P_HEIDI > 0.05, P < 0.05) that passed the heterogeneity test was retained as a robust candidate gene for downstream analysis after cross-validation with the DDR gene set [20]. To evaluate the diagnostic potential of MRE11, we performed paired T-tests on HCC samples from the TCGA database and processed spatial transcriptomic data using the stCancer package. MRE11 expression levels in tumor and adjacent non-tumor tissues were visualized using the HCCDB database (HCCDB: Integrative Molecular Database of Hepatocellular Carcinoma (lifeome.net)). Differences in expression across clinical stages were assessed using unpaired Student’s T-tests, and ANOVA was applied for comparisons between multiple groups. A nomogram was constructed using the rms package to assess the prognostic value of MRE11 in 104 samples, and the Receiver Operating Characteristic (ROC) function of the pROC package was employed to predict 1-, 3-, and 5-year survival rates.

Mutation, epigenetic modification, and regulatory network construction of MRE11
MRE11 mutations were analyzed using the Maftools package, and chi-square tests were employed to evaluate mutation frequency differences. The results were visualized via waterfall plots. We further analyzed the correlation between MRE11 and RNA modification genes, including m1A, m5C, and m6A genes. Copy number variation (CNV) data were obtained from the GDC database for TCGA samples, and integrated with gene expression data. Statistical significance was assessed using Wilcoxon rank-sum tests. DNA methylation analysis was conducted using the UALCAN database [21]. Potential miRNA targets of MRE11 were predicted using four online databases (DIANA-microT [22], miRcode [23], miRWalk [24], and miRDB [25]), and miRNA-lncRNA interactions were analyzed on the StarBase 2.0 platform. Finally, the miRNA-lncRNA-MRE11 regulatory network was visualized using Cytoscape (v2.22.1) and networkD3 (v0.4.1) packages as a Sankey diagram.

Pathway enrichment, immune correlation, and drug sensitivity analysis
Fifty MRE11-interacting proteins were extracted from the STRING database, and 100 MRE11-associated genes were obtained using the GEPIA2 platform for KEGG and GO enrichment analyses to assess their association with the cGAS-STING pathway. Immune infiltration analysis was conducted using the CIBERSORT package, and checkpoint differences among MRE11 expression subgroups were analyzed using the limma package. Drug sensitivity was evaluated using data from TCGA, GDSC (Genomics of Drug Sensitivity in Cancer), and CTRP (The Cancer Therapeutics Response Portal) databases, and the relationship between MRE11 and drugs was visualized via Cytoscape. IC50 (half maximal inhibitory concentration) values were calculated using the pRRophetic package to quantitatively assess drug sensitivity differences between MRE11 expression groups.

Preprocessing and analysis of scRNA-seq data
The scRNA-seq datasets GSE149614, GSE125449, and GSE151530 were analyzed using Seurat v4. Cells were filtered based on criteria of 300 to 7,000 expressed genes, 1,000 to 100,000 unique molecular identifiers (UMIs), and less than 10% mitochondrial genes. The GSE149614 cohort was normalized using the NormalizeData, ScaleData, and FindVariableFeatures functions, selecting the top 3,000 highly variable genes (HVG) to stabilize UMI count variance. Data were integrated using the IntegrateData function, with 2,000 cells selected as anchors. For GSE125449 and GSE151530, normalization and integration were performed using the SCTransform (v0.4.1) function and the Harmony (v1.21) algorithm.

Clustering and cell type annotation
Principal component analysis (PCA) was performed, and the shared nearest neighbor (SNN) graph and uniform manifold approximation and projection (UMAP) were constructed based on the top 30 principal components using the Louvain algorithm. Differentially expressed genes in the GSE149614 dataset were identified using the FindAllMarkers function, and cell types were annotated using established markers [26–30], such as ALB, EPCAM, SERPINA1, and HNF4A for Hepatocytes; CD79A, CD79B, CD37, and MS4A1 for B cells (BCs); CD4, CD3D, CD3E, and IL7R for CD4 + T cells; GZMK, CD8A, and CD8B for CD8 + T cells; GNLY, NKG7, NCR1, and KLRC1 for natural killer cells (NKs); C1orf54, LGALS2, CD1C, and XCR1 for dendritic cells (DCs); CD14, CD163, APOE, C1QA, C1QB, and C1QC for macrophages (Mac); PECAM1, VWF, FLT1, and CLDN5 for endothelial cells (ECs); and COL1A1, DCN, COL1A2, and COL3A1 for fibroblasts (Fibs). For the GSE151530 and GSE125449 datasets, cell types and subpopulations were annotated using SingleR (v2.4.1) and canonical marker-based scoring. The GSE151530 and GSE125449 datasets were annotated using SingleR and marker-based scoring, Additionally, in the GSE151530 dataset, macrophages were further classified into immune-suppressive subpopulations (marked by ‘MRC1’, ‘CD163’, ‘TGFB1’, ‘SPP1’) and immune-activating subpopulations (marked by ‘CD68’, ‘CD86’, ‘TNF’, ‘CD40’) [31–33].

InferCNV (Inference of copy number Variations) analysis
CNV was inferred using the InferCNV (v1.18.1) algorithm. Gene location information was processed, and expression matrices were extracted from the scRNA-seq data, with reference cell groups defined. CNV states in the MRE11 high-expression group were inferred by comparing gene expression levels to the reference cell group, and CNV frequencies were quantitatively analyzed across cell clusters.

Functional enrichment and trajectory analysis of differential genes
Differentially expressed genes between MRE11 high- and low-expression groups were identified using the FindMarkers function, and KEGG and GO pathway enrichment analyses were performed using Gene set enrichment analysis (GSEA) and msigdbr (v1.16.0) analysis. The results were visualized with the ComplexHeatmap package. GSEA was conducted using the clusterProfiler package, and Monocle (v2.30.1) v2 and v3 were used for trajectory inference to evaluate the transition of cells from malignant states to fibroblasts. Dimensionality reduction, classification, and trajectory construction were carried out with default parameters, and significant genes were identified via unsupervised analysis (Benjamini–Hochberg corrected P value < 0.01).

Cell-Cell communication
MRE11-mediated cell-cell communication within the TME of HCC, ligand-receptor interactions were analyzed using the CellChat (v1.5.0) and iTALK (Coolgenome/iTALK (github.com)) packages. Overexpressed genes and ligand-receptor pairs were identified using functions such as identifyOverExpressedGenes, and communication probabilities were calculated. Cell communication networks were constructed using the FindLR function in the iTALK package. Furthermore, the NicheNet (v2.2.0) method was employed to predict ligand-receptor interactions between different cell types in patients receiving immunotherapy, providing further insight into MRE11’s role in regulating the TME.

Spatial transcriptome analysis of MRE11 and cells
In this study, we focused on analyzing the HCC-2T sample [19]. First, we performed co-localization analysis and cell degree analysis of MRE11A + malignant cells and MRE11A + Cancer Associated Fibroblasts (CAFs). Next, we used the CellChat R package to analyze intercellular interactions between the MRE11A + malignant cell subpopulations and CAF subpopulations, specifically focusing on the Hepatocyte Growth Factor (HGF)-Mesenchymal-Epithelial Transition factor (MET) signaling pathway. Finally, the commot package was used to visualize the signaling pathway diagrams of the HGF-MET pathway.

Predictive model construction combining Bulk-seq data and machine learning
To construct an accurate predictive model, we selected 95 candidate genes derived from differential expression in MRE11 high- and low-expression groups in GSE149614, scRNA-seq tumor and adjacent normal tissues, and TCGA-Liver Hepatocellular Carcinoma (LIHC) data. Ten models from a pool of 101 machine learning algorithms, including random survival forest (RSF) (randomForest, v4.7-1.2), Lasso (glmnet, v4.1-8 for Lasso-Cox regression), and Ridge (v3.3), were cross-validated using leave-one-out cross-validation (LOOCV). The model was trained on TCGA-LIHC data and validated using the GSE14520 and ICGC LIRI-JP datasets. The best model was selected based on the Harrell concordance index (C-index), and predictive accuracy was evaluated using time-dependent ROC curves, calibration curves, and decision curve analysis (DCA). Kaplan-Meier (KM) survival analysis was used to assess the model’s prognostic power.

RT-qPCR
Total RNA was extracted from samples using RNAzolt (Huaxingbio, HX16010), followed by reverse transcription into cDNA. Real-time quantitative PCR (RT-qPCR) was performed using the PowerUp SYBR Green Master Mix (Applied Biosystems, 2607157). MRE11 expression was normalized to GAPDH and evaluated using the ΔΔCT method. All experiments were conducted on the ABI 7500 Real-Time PCR System (Applied Biosystems, USA). The primer sequences for MRE11 were: Forward, 5′-GGG TCT CAA AGA GGA AGA GAC AC-3′; Reverse, 5′-GAC ATT TCG GGA AGG CTG CT-3′. GAPDH primers were: Forward, 5′-GGT GGT CTT CTC TGA CTT CAA CA-3′; Reverse, 5′-GTT GCT GTA GCC AAA TTC GTT GT-3′.

Immunohistochemistry (IHC)
HCC tissue samples were fixed with 4% paraformaldehyde and sectioned. Sections were blocked with 5% bovine serum albumin (BSA) at room temperature for 2 h and incubated overnight at 4 °C with an anti-MRE11 recombinant antibody (Huaxingbio, HX16004, 1:150, RRID: AB_3675435). The next day, sections were washed with phosphate buffered solution (PBST) and incubated with goat anti-rabbit secondary antibody (Abcam, ab288151, 1:500, RRID: AB_3675437) at room temperature for 1 h. After washing, the sections were developed using diaminobenzidine (DAB) solution, counterstained with hematoxylin, and observed under a microscope.

Western blotting (WB)
Total protein was extracted from tissue samples using RIPA buffer (Huaxingbio, HX1862-1). Ten micrograms of protein were separated via 10% SDS-PAGE and transferred onto a polyvinylidene fluoride membrane. The membrane was blocked with a 5% BSA blocking buffer (Huaxingbio, HX3303). The blot was incubated overnight with primary antibodies, including rabbit anti-MRE11 (Huaxingbio, HX16004, 1:500, RRID: AB_3675435) and β-tubulin (Huaxingbio, HX1829, 1:5000, RRID: AB_3662663). After three washes with Tris-Borate-Sodium Tween-20 (TBST), the blot was incubated with an anti-rabbit secondary antibody (Huaxingbio, HX2031, 1:10,000, RRID: AB_3572247) for 2 h. Bands were visualized using an Enhanced Chemiluminescence (ECL) detection system (LAS-3000, JAPAN, NO.7612350).

Statistical analysis
Statistical analyses were performed using R software (version 4.3.1) and Python software (version 3.9.11). Parametric tests (Student’s t-test or ANOVA) were applied for normally distributed variables. Non-parametric tests (Wilcoxon rank-sum test or Kruskal-Wallis test) were used to compare continuous variables between two or more groups when the data were non-normally distributed. Pearson’s correlation or Spearman’s rank-order correlation was used to assess relationships between continuous variables. P value < 0.05 was considered statistically significant.

Results
This study analyzed two single-cell RNA-seq datasets and bulk RNA-seq data to explore the role of MRE11 in HCC. Using immune infiltration analysis, pathway enrichment, pseudotime trajectory analysis, spatial transcriptome analysis, and cell-cell communication models, we demonstrated how MRE11 overexpression modulates the immune microenvironment and reshapes the TME. By employing 101 machine learning algorithms, we built a prognostic model based on MRE11-associated genes and validated its robustness as a potential clinical therapeutic target (Fig. 1A).

MRE11 as a potential prognostic target for HCC
Through SMR analysis, we identified 79 genes with potential causal relationships with HCC (Supplementary Table 1). By intersecting these genes with the DDR gene set (Fig. 1B), we identified two candidate genes: MRE11A (P = 0.027) and UVSSA (P = 0.015). Due to the serious heterogeneity observed in UVSSA (P_HEIDI < 0.05), MRE11A (P_HEIDI > 0.05) emerged as the only DDR gene with a causal link to HCC (Fig. 1C). Therefore, we focused further investigations on the MRE11A gene.
MRE11 was found to be significantly upregulated in HCC through paired sample T-tests (P = 1.1e-10) (Fig. 1 D). This finding was further supported by single-cell and spatial transcriptomic analyses, which demonstrated elevated MRE11 expression in HCC tissues (Fig. 1E-F). In a cohort of 92 patients from two major medical centers, qPCR confirmed the differential expression of MRE11 at the RNA level between tumor and adjacent non-tumor tissues (P < 0.0001) (Fig. 1G). IHC staining from 10 patients validated these results at the protein level (P < 0.0001) (Fig. 1H-J). Western blot analysis of tumor and normal tissues from six additional patients corroborated the differential expression of MRE11 (P = 0.03) (Fig. 1K). Stratification of the 92 patients based on MRE11 expression levels (median = 1.139) revealed significant correlations between MRE11 expression and various clinical parameters (Table 1).
Prognostic analyses demonstrated a strong association between MRE11 expression and overall survival (OS, P = 9.8e-3) as well as relapse-free survival (RFS, P = 2.7e-3) (Supplementary Fig. 1 A-B). Kaplan-Meier (K-M) survival analysis further indicated that high MRE11 expression was linked to poorer outcomes in OS (P = 0.0036), RFS (P = 0.015), progression-free survival (PFS, P = 0.0019), and disease-specific survival (DSS, P = 0.0013) (Supplementary Fig. 1 C-F). Correlation analysis revealed significant associations between MRE11 expression and clinical parameters, including TNM stage (P < 0.01), T stage (P < 0.01), tumor grade (P < 0.0001), and patient age (P < 0.001) (Fig. 2A-F, Supplementary Fig. 1 G-J). Based on these findings, we developed a predictive model with a C-index of 0.75 (95% CI: 0.69–0.82). Time-dependent ROC analysis demonstrated area under the curve (AUC) values of 0.77, 0.84, and 0.84 for 1-, 3-, and 5-year survival predictions, respectively, with calibration curves showing high predictive accuracy (Fig. 2G-I, Supplementary Fig. 1 K).

MRE11 mutation analysis and lncRNA-miRNA-MRE11 regulatory network
Mutation analysis revealed that MRE11 is co-mutated most frequently with TP53 (P = 0.02) (Supplementary Fig. 2 A). Mendelian randomization confirmed a positive causal relationship between MRE11 and TP53 (IVW: P = 0.005, OR = 1.13) (Supplementary Fig. 2 B-E). Numerous studies have highlighted the pivotal role of mRNA modifications in cancer progression and incidence [34]. In HCC, MRE11 expression significantly correlated with most mRNA modification genes (P < 0.05) (Supplementary Fig. 2 F), and different modification types—such as m1A, m5C, and m6A—reinforced the association between MRE11 and these genes. These findings suggest that MRE11 may contribute to genomic instability, potentially driving HCC initiation and progression. Copy number variation and DNA methylation analyses suggest that MRE11 expression may be regulated through a complex interplay of epigenetic modifications and other regulatory mechanisms (P < 0.001). The exact mechanisms underlying this regulation remain to be further elucidated in future studies (Supplementary Fig. 2 G-I).
Target miRNAs for MRE11 in HCC were identified using miRDB, DIANA-micro, miRWalk, and miRcode databases, yielding 126, 308, 2108, and 33 miRNAs, respectively. The miRNAs that intersected across two or three databases were identified and individually analyzed using Starbase 2.0, resulting in the identification of eight miRNAs (hsa-miR-30b-5p, hsa-miR-148a-3p, hsa-miR-152-3p, hsa-miR-194-5p, hsa-miR-488-3p, hsa-miR-552-5p, hsa-miR-664b-3p, and hsa-miR-3163), all of which were negatively correlated with MRE11 expression (Supplementary Fig. 3 B-I). These miRNAs were further analyzed for their targeting of lncRNAs, leading to the construction of a comprehensive lncRNA-miRNA-MRE11 regulatory network (Supplementary Fig. 3 A, J, K, Supplementary Table 2). Typically, mRNA expression exhibits a negative correlation with miRNA expression [35], and MRE11 expression was negatively correlated with its target miRNAs. Consistent with the ceRNA hypothesis, lncRNAs likely act as competing endogenous RNAs (ceRNAs), sponging these miRNAs to regulate MRE11 expression [36]. This highlights a dual regulatory mechanism where lncRNAs and miRNAs intricately control MRE11 expression, providing potential insights into post-transcriptional regulation in HCC.

Pathway enrichment, immune correlation, and drug sensitivity analysis
KEGG and GO enrichment analyses revealed that MRE11-related genes are closely associated with DDR pathways (Fig. 3A, Supplementary Fig. 4 A-B). Notably, MSH2 emerged as the sole overlapping gene between the 50 MRE11-associated interactome and 100 strongly related genes of MRE11. As a core DNA mismatch repair component, its co-occurrence with MRE11 suggests potential epistatic interactions or convergent pathways driving genomic instability in MRE11-high HCC patients (Supplementary Fig. 4 A-B). In addition, given the close relationship between the cGAS-STING pathway and DDR [6], we analyzed the correlation between MRE11 and the cGAS-STING pathway (P = 1.9e-21, R = 0.47), as well as its associated genes (P < 0.01), revealing a significant association (Fig. 3B-C, Supplementary Fig. 4 C-J). Since cGAS-STING pathway activation mediates immune responses, we conducted an immune infiltration analysis. High MRE11 expression correlated with the activation of various immune cell types, including macrophages (P < 0.05), T cells (P < 0.05), dendritic cells (P < 0.01), and B cells (P < 0.001) (Fig. 3D, Supplementary Figs. 4 K-L).

Further analysis revealed significant positive correlations between high MRE11 expression and immune checkpoints, including PD-1 (P < 0.001), PD-L1 (P < 0.01), CTLA4 (P < 0.001), and HAVCR2 (P < 0.001) (Fig. 3G). Subsequently, we performed a drug sensitivity analysis of the MRE11 gene using the GDSC and CTRP databases (Supplementary Table 3), presenting the top 30 drugs with positive and negative correlations (Fig. 3E-F). We also quantified IC50 values for the high and low expression groups of MRE11 (Supplementary Table 3), visualizing results for 20 drugs (Supplementary Figs. 5 A-T).

MRE11-Driven heterogeneity in the tumor microenvironment
After quality control and filtering, we obtained 58,471 high-quality single-cell transcriptomes from the GSE149614 dataset, comprising 27,443 non-tumor and 31,028 HCC cells. Dimensionality reduction and clustering revealed key heterogeneity across different tissue types, hepatitis types, and pathological stages. Major cell populations identified include Hepatocytes, BCs, CD4 + T, CD8 + T, NKs, DCs, Mac, ECs, and Fibs (Fig. 4A-C, Supplementary Fig. 6 A-F). We analyzed cellular composition at the individual sample level, noting distinct differences in cell origins between normal and cancerous tissues (Fig. 4D, Supplementary Fig. 6 B, D). In HCC tissues, malignant hepatocytes predominated, while immune cell infiltration was more prominent in normal tissues. Tumor progression correlated closely with cancer transformation, showing increased infiltration of immune cells with advancing tumor stages (Supplementary Fig. 6 C-D). Additionally, hepatocytes were more abundant in HCV patients than in HBV and non-hepatitis patients, whereas B and T cell infiltration was lower (Supplementary Fig. 6 E-F).

To further investigate the heterogeneity of the tumor microenvironment related to MRE11, we compared the cellular compartments between high and low expression subgroups of MRE11 (Fig. 4E). The MRE11-high group exhibited higher hepatocyte content and lower levels of immune cells. The Infer CNV analysis of malignant cells (n = 12,001) revealed increased copy number variations (CNVs) and a higher degree of malignancy in the MRE11-high group, with a similar level of heterogeneity observed among clusters (Supplementary Fig. 6 G-H).

MRE11-Driven ecosystem changes in the HCC tumor microenvironment
Single-cell resolution analysis revealed significant variation in MRE11 expression across different cell types (Fig. 4F). Interestingly, MRE11 expression was higher in T cells, NK cells, hepatocytes, and macrophages. Differential expression analysis between MRE11 high and low groups identified 371 differentially expressed genes (Fig. 4G). KEGG and GO enrichment analyses linked these genes to the cGAS-STING pathway. AUC scores confirmed the enrichment of differential genes and the cGAS-STING pathway in hepatocytes and immune cells (Fig. 4H, J). KEGG and GO analysis showed that the cGAS-STING pathway was significantly upregulated in the high MRE11 expression group, exhibiting negative regulation in malignant epithelial cells and positive regulation in immune cells (Fig. 4I-K). GSEA analysis revealed suppressed immune responses in the high MRE11 expression group, along with the activation of pro-tumor pathways (Liver_Cancer_Krt19_Dn pathway [37]), indicating that MRE11 may promote tumorigenesis in HCC by modulating the immune ecosystem (Fig. 5A-B, Supplementary Table 4).

To explore MRE11’s impact on the TME, pseudotime analysis revealed that high MRE11 expression may drive hepatocytes towards a fibroblast-like phenotype, leading to a fibrotic TME and reduced drug sensitivity (Fig. 5C). Temporal trajectory analysis indicated MRE11 expression showed a decreasing trend over time, and the differential genes exhibited two opposite expression trends, further confirming MRE11 dynamic role in HCC (Fig. 5D-E, Supplementary Fig. 6 I-J). To explore the specific mechanism of MRE11-mediated fibrosis in hepatocytes, we used CellChat and iTALK to compare cellular interactions between MRE11 high- and low-expression groups. Hepatocytes in the MRE11-high group showed significantly more frequent interactions with fibroblasts, endothelial cells, macrophages, and CD4 + T cells, whereas there was no significant difference in interactions between other cell types (Fig. 5F-G, Supplementary Fig. 7 A-H). Ligand-receptor analysis revealed that fibroblast-hepatocyte interactions were most prominent, with notable activation of the HGF-MET pathway in the MRE11-high group, suggesting a key role in liver fibrosis (Fig. 5H-I, Supplementary Fig. 7 I). The HGF-MET signaling pathway plays a crucial role in liver fibrosis, where the upregulation of HGF expression triggers the activation, proliferation, and migration of fibroblasts through the activation of MET receptors [38]. Finally, by iTALK analysis, we further refined the immune checkpoint-, cytokine-, and growth factor-associated ligand–receptor interactions in the MRE11 high-expression group, revealing their important role in shaping the TME (Supplementary Fig. 7 J-M). These results suggest that MRE11 not only plays a role by regulating the tumor immune microenvironment, but may also contribute to the development and progression of HCC by enhancing intercellular communication, especially in the process of immune escape and fibrosis.

MRE11 enhances malignant Cell-CAF interaction via HGF-MET pathway
Using UMAP dimensionality reduction, we identified cell populations within the GSE125449 dataset, including malignant cells, HPC-like cells (normal hepatocytes), B cells, CD4 + T cells, NKT cells, TAMs (Tumor-associated Macrophages), ECs, and CAFs (Fig. 6A). We then examined the differential expression of MRE11 between malignant and HPC-like cells, finding it significantly upregulated in malignant cells, suggesting that it may play a potential role in tumorigenesis. Pseudotime analysis showed a bifurcation in hepatocyte differentiation—toward either normal or malignant states—supporting MRE11’s involvement in malignant transformation (Fig. 6B-D).

CellChat analysis revealed enhanced communication between MRE11-positive malignant cells and CAFs (Fig. 6E). Spatial transcriptomics further showed that MRE11 + malignant cells were primarily localized in tumor regions, particularly near CAF-rich areas (Fig. 6F-G). By analyzing the intercellular communication of the HGF-MET signaling pathway, we detailed the interaction direction between HGF (ligand) and MET (receptor) among malignant cells and CAFs. The results showed highly active signaling of the HGF-MET pathway between MRE11 + malignant cells and CAFs, further verifying the key role of MRE11 in regulating the HGF-MET pathway (Fig. 6H-J). Meanwhile, spatial analysis confirmed significant co-localization of MRE11 + CAFs and malignant cells (Fig. 6K), while Fig. 6L demonstrated stronger interactions between MRE11 + CAFs and malignant cells.

MRE11 regulation of the TME Post-Immunotherapy
To further investigate the role of MRE11 in immunotherapy, we analyzed the GSE151530 single-cell dataset, which includes 30,458 cells from HCC tissues pre- and post-immunotherapy, comprising 10,856 cells from untreated tissues and 19,602 cells from treated tissues (Fig. 7A). After data integration and batch effect correction (Supplementary Fig. 8 A), the major cell populations identified included malignant cells, B cells, CD4 + T cells, CD8 + T cells, NK cells, TAMs, Tumor Endothelial Cells (TECs), and CAFs (Fig. 7A). Post-immunotherapy, the proportion of malignant cells significantly decreased, while immune cell infiltration increased, particularly in the MRE11 high-expression group, indicating a pivotal role for MRE11 in modulating immunotherapy response (Fig. 7B-D). Conversely, in patients who did not receive immunotherapy, the MRE11 high-expression group exhibited a higher proportion of malignant cells and reduced immune cell infiltration (Fig. 7B-D). MRE11 was broadly expressed in both malignant and immune cells in the immunotherapy group, suggesting its potential regulatory role in reshaping the tumor microenvironment (Fig. 7E). Pathway enrichment analysis revealed significant enrichment of the cGAS-STING pathway in various immune cells—including macrophages, B cells, NK cells, CD4 + T cells, and CD8 + T cells—in the MRE11 high-expression group (Fig. 7F). Further enrichment analysis confirmed that this pathway was markedly upregulated in TAMs and NK cells but downregulated in malignant cells and CAF subpopulations (Fig. 7G-H). Notably, we also observed a significant downregulation of the PPAR signaling pathway [39] (Supplementary Fig. 8 B, Supplementary Table 5).Previous studies have implicated this pathway in angiogenesis and metabolic regulation in cancer, yet its precise role in tumor growth and progression remains controversial [40, 41]. Collectively, these findings suggest that MRE11 may influence antitumor immune responses by dynamically modulating immune regulation and tumor metabolic pathways.

Monocle v3 analysis revealed that hepatic malignant cells and endothelial cells partially differentiated into mature malignant cells and CAFs (Fig. 8A), suggesting that MRE11 may drive tumor progression and drug resistance through promoting fibrosis and immune evasion. Pseudo-time trajectory analysis demonstrated dynamic changes in differential gene expression during tumor evolution, further indicating the role of MRE11 in regulating the HCC immune microenvironment (Supplementary Fig. 8 C). Given the dynamic reprogramming of TME cellular communication in HCC, we employed CellChat, NicheNet, and iTALK tools to dissect cell-to-cell communication in the MRE11 high-expression group. The analysis showed that ligand-receptor interactions between malignant cells, endothelial cells, and macrophages were significantly enhanced in the MRE11 high-expression group (Fig. 8B-E, Supplementary Fig. 8D-I). NicheNet analysis revealed that APOE-LRP1 ligand-receptor pairs were particularly active in malignant cell-macrophage interactions, while A2M-TNFRSF14 pairs were prominent in the endothelial cell-macrophage axis (Fig. 8F-G). MRE11 high expression significantly alters the ligand-receptor networks of cytokines, immune checkpoints, growth factors, and other factors, especially in the immune therapy group. Enhanced interactions between macrophages and other immune cells suggest that MRE11 is crucial in tumor immune responses by modulating intercellular communication within the TME (Fig. 8H, Supplementary Fig. 8J-L). Subpopulation analysis of macrophages indicates a higher proportion of immunosuppressive macrophages in the MRE11 high-expression group compared to the low-expression group. Pseudotime analysis reveals an increased differentiation of monocytes and granulocyte-monocyte progenitors (GMP) into immunosuppressive macrophages (Figs. 8I-K). These results underscore MRE11’s critical role in the polarization of tumor-associated macrophages (TAMs) toward an immunosuppressive phenotype.

Identification of differential genes and construction of a robust prognostic model
After establishing MRE11 as a poor prognostic indicator for HCC, we further utilized the identified transcriptional features by selecting 95 overlapping genes from scRNA-seq and bulk-seq datasets (Fig. 9A). A LOOCV framework was used to fit 101 machine learning models, with the C-index calculated for each model in both training and validation cohorts. The optimal predictive model identified was a combination of RSF and Ridge regression, with an average C-index of 0.71 (Fig. 9B, Supplementary Fig. 9 A). RSF importance analysis identified 61 differential genes when ntree = 800 (Fig. 9C-D). The optimal lambda value (Log λ = 0.46) was determined using the LOOCV framework when the partial likelihood deviance reached its minimum (Fig. 9E-F). Rigorous internal and external validation of the model was conducted. K-M survival analysis of the TCGA-LIHC and GSE14520 datasets showed that patients in the high-risk group, as classified by model-derived risk scores, had significantly lower OS than those in the low-risk group (P < 0.0001, Fig. 9G-H).

Time-dependent ROC (Fig. 9I-N) and Time-dependent C-index analyses (Supplementary Fig. 9 B-C) confirmed the model’s predictive performance, with AUC and C-index values exceeding 0.7 in both the training and validation cohorts. Calibration and DCA demonstrated good model fit and accuracy (Supplementary Fig. 9 D-K). These results confirmed the robust predictive performance of the 61 genes identified from the MRE11 differential gene screening, suggesting that the constructed model may serve as a potential prognostic tool for HCC patients.

Discussion
This study, using SMR analysis, identified MRE11A as the only DDR gene with a causal association with HCC. Through multi-omics analysis and experimental validation, we comprehensively investigated MRE11’s key role in HCC and its potential prognostic value. MRE11 was significantly upregulated in HCC, as confirmed by TCGA, single-cell, and spatial transcriptomics. Validation through qPCR, immunohistochemistry, and Western blot confirmed its differential expression at both RNA and protein levels. Survival analysis indicated a strong correlation between MRE11 and several prognostic factors (OS, RFS, PFS, DSS), underscoring its role in HCC progression and patient survival. A predictive model based on MRE11, TNM stage, T stage, and age achieved high accuracy, with AUC values exceeding 0.75 for 1-, 3-, and 5-year survival, reinforcing MRE11 as a reliable prognostic marker. Mutation and Mendelian randomization analyses revealed a positive causal relationship between MRE11 and TP53, with MRE11 exhibiting a high mutation frequency linked to HCC progression. We also constructed a lncRNA-miRNA-MRE11 regulatory network in HCC. Eight miRNAs, identified from multiple prediction tools, target both MRE11 and lncRNAs, negatively correlating with MRE11 expression. These miRNAs likely regulate MRE11 through lncRNA-mediated miRNA sponging, supporting the ceRNA hypothesis. This dual-layered regulation underscores MRE11’s complex control in HCC. The identified miRNAs and lncRNAs are potential therapeutic targets, warranting further validation. Pathway enrichment analysis linked MRE11 to DDR and the cGAS-STING pathway, both critical in tumor immune response and surveillance [1, 3, 10]. Immune infiltration analysis showed that high MRE11 expression activated several immune cell types and positively correlated with immune checkpoints (PD-1, PD-L1, and CTLA-4), suggesting involvement in immune evasion. This highlights MRE11 as a potential therapeutic target in HCC. Drug sensitivity analysis further indicated differential responses between high and low MRE11 expression groups, emphasizing its potential in personalized HCC treatment.
Through single-cell sequencing analysis, we further explored MRE11 heterogeneity within the HCC TME. We found that high MRE11 expression was associated with a higher proportion of malignant hepatocytes, indicating a strong correlation between MRE11 overexpression and tumor cell proliferation and expansion. Moreover, high MRE11 expression is significantly associated with a lower proportion of immune cells, suggesting that it may promote tumor immune evasion by regulating immune cell infiltration and function. InferCNV analysis confirmed higher frequencies of CNVs in the high MRE11 group, suggesting that MRE11 contributes to genomic instability, further driving malignant transformation in HCC. Differential gene expression and KEGG/GO enrichment analyses revealed that MRE11 negatively regulates cGAS-STING pathway in malignant epithelial cells, while it positively regulates this pathway in immune cells. Pseudotime trajectory analysis uncovered the potential role of high MRE11 expression in promoting the transition of tumor cells to fibroblasts, suggesting its involvement in tumor fibrosis. Cell-cell communication analysis further demonstrated enhanced interactions between hepatocytes and macrophages, endothelial cells, and fibroblasts in the high MRE11 expression group, particularly through the HGF-MET ligand-receptor axis, implicating MRE11 in liver fibrosis and tumor invasion.
Utilizing the GSE125449 dataset, we examined the role of MRE11 in the communication between malignant cells and CAFs. High MRE11 expression in malignant cells is crucial for the malignant transformation of hepatocytes. Pseudotime analysis revealed a bidirectional differentiation trajectory from hepatocytes to malignant cells, reinforcing MRE11’s role as a driving factor. CellChat analysis showed significantly enhanced communication between MRE11 + malignant cells and CAFs, particularly through the HGF-MET signaling pathway, underscoring MRE11’s regulatory influence on their interactions. Spatial transcriptomics further confirmed the colocalization of MRE11 + malignant cells and CAFs in the tumor center, indicating that MRE11 promotes their interaction by activating the HGF-MET pathway. Previous studies have shown that CAFs mediate the HGF/MET signaling pathway to enhance cancer cell invasion and immune evasion [42]. However, whether MRE11 expression regulates the HGF/MET pathway in modulating tumor-CAF interactions in HCC remains unreported. Further studies are required to validate this potential mechanism. Collectively, these results support that high MRE11 expression reprograms the immune ecosystem, particularly by regulating the cGAS-STING pathway and the HGF-MET axis, driving HCC progression from DNA repair dysfunction to malignancy.
In the context of immunotherapy, we observed significant changes in the proportion of malignant and immune cells among patients receiving immunotherapy compared to those who did not. Notably, in the immunotherapy group, the ratio of malignant cells to immune cells shifted markedly in the MRE11 high-expression subgroup, suggesting that MRE11 may play a pivotal role in modulating tumor response to immunotherapy. MRE11 was highly expressed in tumor cells and downregulated immune-related pathways, notably the cGAS-STING pathway, in malignant cells. Interestingly, this pathway was upregulated in immune cells, particularly TAMs, highlighting MRE11’s dual regulatory role in immune responses. These findings indicate MRE11’s potential for shaping the tumor immune microenvironment by suppressing immune pathways in malignant cells while promoting immune activation in TAMs. Furthermore, single-cell trajectory analysis revealed that in the high MRE11 group, malignant cells differentiated into both mature malignant cells and fibroblasts, while endothelial cells differentiated into fibroblasts, further promoting tumor proliferation, metastasis, and drug resistance. Using tools such as CellChat and iTALK, we found that the interaction between malignant cells and macrophages was notably active, especially through the APOE-LRP1 [43] and A2M-TNFRSF14 [44, 45] ligand-receptor pairs, suggesting that MRE11 regulates immune response and tumor progression through these axes. Subpopulation analysis of macrophages revealed a higher proportion of immunosuppressive macrophages in the high MRE11 group, with pseudotime analysis further confirming the differentiation of monocytes and GMPs into immunosuppressive macrophages.
After establishing MRE11 as a clear predictor of poor prognosis in HCC, we developed a robust prognostic model for HCC based on the differentially expressed genes associated with MRE11. We screened 95 differently genes from single-cell RNA-seq and bulk-seq datasets and applied 101 machine-learning models to validate key genes associated with HCC prognosis. Using a combination of RSF and Ridge regression, we successfully developed a prediction model based on the LOOCV framework, which exhibited strong performance across multiple independent datasets. Internal and external validation using K-M survival analysis, time-dependent ROC curves, and C-index analyses, along with calibration and decision curves, confirmed the model’s accuracy and robustness. These results suggest that the 61 genes identified from the high- and low-expression MRE11 subgroups exhibit strong predictive accuracy, and the developed prognostic model may serve as a valuable tool for forecasting outcomes in HCC patients.
Despite the comprehensive multi-omics analysis and experimental validation presented in this study, several limitations remain. First, although we demonstrated the significant role of MRE11 in HCC progression and immune regulation, the underlying molecular mechanisms, particularly its interaction with the cGAS-STING and HGF-MET pathways, require further in-depth functional studies. Second, while single-cell and spatial transcriptomics provided valuable insights into MRE11 heterogeneity within the tumor microenvironment, validation in larger, independent cohorts is needed to confirm its clinical relevance. Finally, the immunosuppressive role of MRE11 in HCC immunotherapy warrants further investigation using preclinical models and clinical samples to elucidate its potential as a therapeutic target. Future studies integrating advanced techniques, such as CRISPR-Cas9 and organoid models, will help address these limitations and strengthen the translational value of our findings.
By integrating external database analyses with internal experimental validation, this study systematically elucidates the critical role of the MRE11 gene in HCC progression and treatment response. Our research highlights that high MRE11 expression significantly reshapes the HCC TME, especially in primary tumors and under immunotherapy, underscoring its central role in tumor immune escape and therapeutic response. Furthermore, the 61 differentially expressed genes associated with MRE11, identified from the high- and low-expression subgroups, serve as reliable prognostic biomarkers for HCC and enable a robust predictive model for patient outcomes. In the future, combining multi-omics approaches, MRE11 holds promise as a potential target for HCC diagnosis and personalized therapy, laying the foundation for targeted treatment strategies against MRE11.

Supplementary Information