Abstract
BACKGROUND: Esophageal squamous cell carcinoma (ESCC) remains a major malignancy globally, and long-term survival outcomes remain poor despite advances in multimodal treatment strategies. Neoadjuvant therapy (NAT) has become the standard of care for resectable ESCC; however, substantial interpatient heterogeneity in treatment response persists. Defining malignant cell-intrinsic molecular determinants linked to NAT response is essential for improving prognostic stratification and informing individualized therapeutic decision-making. METHODS: To investigate transcriptional alterations in malignant epithelial cells, we analyzed scRNA-seq datasets obtained from ESCC patients both before and after NAT. Overlapping molecular features were identified by integrating differentially expressed genes from pre- and post-treatment malignant cells with bulk RNA-seq data from TCGA and GEO cohorts. Functional enrichment analyses, pseudotime trajectory reconstruction, and transcription factor regulatory network assessments were subsequently performed. Candidate prognostic genes were initially screened through univariate Cox analysis, followed by the development of the NTAMPS model using LASSO-Cox regression. Its prognostic performance was validated in the TCGA-ESCA and GSE53624 cohorts, and clinical applicability was further examined using a nomogram, calibration curves, and decision curve analysis. Immune-related associations, immunotherapy response prediction (IMvigor210 and GSE78220), and drug sensitivity profiling were also conducted. DUXAP8, which carried the largest coefficient in the NTAMPS, was selected for further validation through both in vivo and in vitro assays, including qRT-PCR and evaluations of cell proliferation, migration, invasion, and colony formation. RESULTS: Malignant epithelial cells exhibited pronounced transcriptional remodeling after NAT, characterized by enrichment of pathways related to the cell cycle, DNA replication, and epithelial-mesenchymal transition. Cell-cell communication analysis revealed substantial reorganization of the MIF signaling pathway, including increased interactions of MIF-ACKR3 and MIF-(CD74+CXCR4/CD44), with fibroblasts acting as major signal senders and macrophages serving as primary receivers. Twenty-one prognosis-related genes were identified, and a ten-gene NTAMPS demonstrated strong prognostic performance. Both NTAMPS and clinical stage emerged as independent prognostic factors and were integrated into a nomogram with favorable calibration and decision curve characteristics. A high NTAMPS was associated with an immunosuppressive microenvironment, reduced predicted response to immunotherapy, and distinct drug sensitivity patterns. DUXAP8, the top positive risk gene within NTAMPS, was highly expressed in ESCC tissues and cell lines, and its silencing suppressed cell proliferation, migration, invasion, and clonogenic potential. CONCLUSION: This study establishes NTAMPS as a novel malignant cell-derived prognostic signature for ESCC by integrating single-cell and bulk transcriptomic data. NTAMPS enables effective prognostic stratification, predicts potential immunotherapy benefit, and highlights therapeutic vulnerabilities. DUXAP8 was further identified as a candidate molecular driver that may improve ESCC management in the context of neoadjuvant therapy.