Abstract
<title>Abstract</title> <p>Background Hepatocellular carcinoma (HCC) exhibits pronounced heterogeneity. Existing prognostic models often lack an immunological perspective. This study leveraged single-cell transcriptomics to characterize T-cell features, constructing and validating a biologically interpretable prognostic model to enhance precise risk stratification in HCC. Methods Single-cell RNA-seq data from 6 HCC patients were analyzed to identify T cell-specific genes. A 6-gene prognostic model was then constructed using differential expression and Cox/LASSO regression in the TCGA-LIHC, and validated in GSE76427. Model performance was evaluated via survival analysis, ROC curves, C-index and decision curve analysis, and a clinical nomogram was developed. Further analyses characterized the immune microenvironment, functional pathways, and immunotherapy relevance (CIBERSORT, ESTIMATE, TIDE). Finally, the key gene SLC4A10 was experimentally validated using qRT-PCR and functional assays in vitro. Results A total of 751 T-cell-related DEGs were identified in HCC, primarily enriched in immune response and T-cell activation pathways. The final 6-gene risk model significantly stratified patients into high-risk and -risk groups in the TCGA (log-rank p < 0.001), achieving a C-index of 0.709. The area under the ROC curve (AUC) values for 1-year, 3-year and 5-year overall survival were 0.790, 0.740 and 0.750, respectively. The risk score was an independent prognostic factor (p < 0.001). The nomogram demonstrated enhanced predictive efficacy (C-index = 0.78). In the GSE76427, the model successfully stratified patients (log-rank p = 0.021) and outperformed several published models. The high-risk group exhibited enrichment in pro-proliferative pathways. Their immune microenvironment was characterized by elevated TIDE scores, restricted T-cell infiltration, enrichment of MDSCs and Tregs, and high expression of PD-L1 and multiple immune checkpoints. Conversely, the low-risk group had higher IPS scores, suggesting greater potential responsiveness to immunotherapy. Although TMB was higher in the high-risk group, the risk score remained an independent prognostic factor regardless of TMB status. Experimental validation confirmed higher SLC4A10 expression in adjacent non-tumor tissues. Low expression promoted HCC cell proliferation, migration and clonogenic growth in vitro, collectively supporting its function as a tumor suppressor gene. Conclusion This study developed a prognostic model based on single-cell T-cell signatures for HCC, enabling effective risk stratification. It delineates distinct immune and metabolic landscapes between risk groups, offering novel insights into tumor biology and nominating SLC4A10 as a potential therapeutic target.</p>