Malignant epithelial cell marker-driven risk signature enables precise stratification in esophageal cancer

Esophageal squamous cell carcinoma (ESCC) is a highly heterogeneous malignancy. Traditional clinical staging systems have limited value in prognostic evaluation and treatment guidance, often failing to capture the profound impact of intratumoral diversity on patient outcomes. Single-cell RNA sequencing (scRNA-seq) provides new perspectives on the cellular makeup and conditions in tumor tissues, with promising implications for functional classification and personalized therapeutic strategies in esophageal cancer.

In this study, we integrated scRNA-seq data with bulk RNA-seq profiles from esophageal cancer tissues to construct a comprehensive cellular atlas, focusing on the transcriptional characteristics of epithelial cells. Malignant epithelial cells were identified based on copy number variation (CNV) features using inferCNV analysis. Their developmental states and regulatory mechanisms were further characterized via transcription factor activity inference (SCENIC) and pseudotime trajectory analysis (Monocle). Based on marker genes of malignant epithelial subpopulations, we developed a multi-gene risk scoring model using data from the TCGA and GEO (GSE53624) cohorts. The model's predictive value for immune landscape, mutational features, and drug sensitivity was also evaluated. Additionally, qRT-PCR was conducted to quantify the expression levels of model genes in ESCC tissue samples, further evaluating their biological relevance. Functional roles of the key gene HMGB3 were validated in vitro through CCK-8 proliferation assays, Transwell invasion assays, and colony formation assays following gene knockdown in ESCC cell lines.

At the single-cell level, we identified ten major cell types and six distinct malignant epithelial subclusters, which exhibited pronounced heterogeneity in cell cycle states, transcriptional regulatory networks, and differentiation trajectories. High CNV scores and the enrichment of specific transcription factors (e.g., FOXC1, E2F1, RUNX1) suggested a proliferative and immune-evasive phenotype. A six-gene prognostic model (HMGB3, CHORDC1, CTSD, BTG2, MT1E, PHYHD1) showed strong predictive accuracy for overall survival in the TCGA and GSE53624 cohorts. Furthermore, the risk score showed a significant correlation with an immunosuppressive tumor microenvironment, increased tumor purity, and the activation of certain immune-related pathways. Analysis of drug sensitivity suggests that patients classified as low-risk could respond better to various targeted therapies and chemotherapeutic agents, underscoring their potential clinical relevance. Functional assays revealed that HMGB3 knockdown markedly suppressed ESCC cell proliferation, invasion, and colony formation, supporting its oncogenic role in esophageal cancer progression.

This study systematically characterized epithelial cell heterogeneity in esophageal cancer at single-cell resolution and established a risk model based on malignant epithelial markers that effectively predicts prognosis, immune status, and potential drug response. Combined with experimental validation, our findings highlight the pivotal role of HMGB3 in ESCC progression and provide a novel theoretical and practical framework for functional tumor classification and individualized treatment strategies.