Recent advances in single-cell sequencing have revolutionized our understanding of gene expression profiles at the individual cell level, providing groundbreaking insights for precision cancer therapy [20, 21]. High-throughput sequencing of both tumor and immune cells offers new perspectives for examining the tumor immune microenvironment, the occurrence and metastasis of hepatocellular carcinoma, and aspects including clinical diagnosis, individual therapeutic differences, prognosis, and assessment [22]. Hepatocellular carcinoma is marked by significant cellular heterogeneity within tumor tissues. Single-cell sequencing is instrumental in identifying biomarkers, particularly in understanding the role of T cells in the TME [22, 23], which could enhance the precision treatment of hepatocellular carcinoma. The scRNA-seq dataset used in this study, obtained from the GEO database, facilitated a unique analysis, distinct from previous research. This study focused on elucidating the specific roles of T-cell marker genes in the tumor microenvironment and leveraged their expression profiles to develop a prognostic risk model aimed at improving HCC patient outcomes. A T-cell prognostic risk model, based on the LIHC dataset from TCGA and centered on T-cell marker genes in hepatocellular carcinoma, was established. The model's reliability was validated using a dataset from the GEO database. Eight genes of prognostic importance were identified: PTTG1, STMN1, UBE2S, RTKN2, S100A10, CITED2, SLC38A1, and CD69.

Studies have demonstrated that high expression of PTTG1, the pituitary tumor transforming gene, is associated with increased T cell permeability in tumors [24, 25]. PTTG1 also plays a role in tumorigenesis by upregulating the transcription of asparagine synthetase, a crucial enzyme in polyamine biosynthesis, with T-cell infiltration closely linked to immune surveillance and the efficacy of immunotherapy [24]. T cells are essential in tumor immune responses, and PTTG1 may serve as a predictive biomarker for immune checkpoint blockade (ICB) responses. STMN1, a cytoplasmic phosphoprotein involved in microtubule dynamics, has been linked to the prognosis of various malignancies [26, 27]. Rui et al. [26] found that STMN1 mediates interactions between hepatocellular carcinoma cells and hepatic stellate cells through the Hepatocyte Growth Factor/MET signaling pathway, influencing tumor behavior and prognosis. Ubiquitination plays a critical role in LIHC progression, affecting processes such as protein degradation and signal transduction. The ubiquitin-conjugating enzyme E2S (UBE2S), a member of the ubiquitin-binding enzyme family, is associated with tumor size, stage, and TNM classification. UBE2S promotes AKT phosphorylation and reduces PTEN protein levels by facilitating PTEN ubiquitination at Lys60 and Lys327, impacting immune responses [28, 29]. S100A10, part of the S100 protein family, is involved in converting fibrinogen to an active protease [30]. Its upregulation in various cancers suggests its potential as a biomarker for cancer prognosis [31, 32]. SLC38A1, a glutamine transporter, plays a significant role in tumor cell migration and transport [33]. Yun Liu et al. [33] reported that increased SLC38A1 expression correlates with hepatocellular carcinoma prognosis and immune infiltration, corroborating our findings.

Rhotekin 2 (RTKN2) is recognized as an oncogene in locally advanced gastric cancer (GC). Although RTKN2 overexpression is known to enhance GC cell proliferation and migration, its specific role in liver hepatocellular carcinoma (LIHC) remains unexplored, suggesting a potential, yet unidentified, function in LIHC development [34]. Additionally, Zhao HG et al. [34] reported that RTKN2 significantly influences the Wnt/β-catenin signaling pathway. Hypoxia, a common feature in various cancers, involves crucial roles in the induction, activation, and stabilization of hypoxia-inducible factor 1 (HIF-1) in cancer's biological functions. CITED2, widely expressed, binds competitively to the CH1 domain of CBP/p300, a transcriptional co-activator, counteracting HIF-1, which is essential for tumor cell adaptation to hypoxic environments [35]. CITED2 functions as both a positive and negative regulator [28]. Molecular studies reveal that CITED2 deficiency results in increased IFN-γ-induced STAT1 transcriptional activity and enhanced STAT1 presence in macrophages. Atif Zafar et al. [36] demonstrated that CITED2 inhibits the STAT1-IRF1 signaling pathway in macrophages, reducing plaque formation in atherosclerosis. The roles of CITED2 and RTKN2 in predicting LIHC outcomes are yet to be thoroughly investigated. Given their established roles in other tumors, these genes could serve as novel biomarkers for LIHC prognosis, meriting further study.

In this study, we successfully constructed a prognostic risk model for HCC based on the TCGA-LIHC dataset and validated its robustness and predictive performance using multiple independent datasets, including GSE76427 and ICGC. One of the most significant findings is that the low-risk group exhibited significantly better overall survival (OS) than the high-risk group. The eight independent prognostic DEGs identified through multivariate Cox regression showed excellent predictive power in the TCGA cohort, with AUCs of 0.8, 0.76, and 0.76 for 1-year, 3-year, and 5-year OS prediction, respectively. These results were further validated in the GSE76427 and ICGC datasets, confirming the model's robustness and generalizability.

Compared to other hepatocellular carcinoma (HCC) prognostic models, our model offers notable advantages. For instance, the gene expression signature model developed by Hoshida et al. [37], while proficient in predicting HCC prognosis, yielded a moderate area under the curve (AUC) of approximately 0.74 in external validation datasets, indicating limited predictive accuracy. Conversely, our model achieved higher AUC values for overall survival (OS) at 1-year, 3-year, and 5-year intervals, with AUCs of 0.8, 0.76, and 0.76, respectively, suggesting an enhanced predictive capability for HCC prognosis. Furthermore, the immune-related gene signature model by Zhang et al. [38] reported AUCs of 0.73, 0.72, and 0.71 for predicting 1-year, 3-year, and 5-year OS in the TCGA validation cohort, highlighting the potential of immune genes in HCC prognosis. Zhou et al.'s [39] study, "T Cell-based Prognostic Prediction of Liver Cancer Patients," introduced a prognostic model focused on T cell depletion, achieving AUCs of 0.82, 0.75, and 0.72 for 1-year, 3-year, and 5-year OS, respectively. Despite the moderate predictive accuracy of their model, its limited validation in independent datasets suggests reduced generalizability.

Overall, our model not only demonstrates high predictive accuracy in the TCGA cohort but also maintains its robustness and generalizability across multiple independent datasets. Compared with other models, our model shows superior performance in long-term survival prediction, providing a new tool and direction for early diagnosis and personalized treatment of HCC patients.

These findings carry profound clinical implications. By pinpointing gene features intimately linked to patient prognosis, we have equipped clinicians with more precise tools for decision-making, thereby facilitating the identification of high-risk patients and the formulation of tailored treatment strategies. Specifically, the elevated expression of genes such as PTTG1, STMN1, and UBE2S is strongly correlated with unfavorable prognosis, whereas the expression of genes like CITED2 and SLC38A1 in the low-risk group is associated with more favorable outcomes. These genes are intricately involved in a myriad of biological processes, including cell cycle regulation, protein degradation, and metabolic regulation, underscoring their potential roles as pivotal drivers in the pathogenesis and progression of hepatocellular carcinoma.

Exhausted T cells are generally more prevalent in the tumor microenvironment due to persistent antigen stimulation and the presence of an immunosuppressive milieu. However, we observed a higher proportion of exhausted T cells in normal tissue compared to tumor tissue, indicating an underlying mechanism of immune balance. In normal tissues, chronic exposure to antigens, such as those from microbial or environmental sources, can lead to gradual T cell exhaustion, serving to maintain immune homeostasis and prevent excessive immune responses. Research has demonstrated that chronic antigen exposure in non-tumor environments can induce T cell exhaustion as a regulatory mechanism [40, 41].

Conversely, within the tumor microenvironment, tumor cells can secrete immunosuppressive molecules and modulate antigen presentation, thereby inhibiting the accumulation of exhausted T cells while maintaining a certain proportion of effector T cells. This may be a strategy that tumor cells use to evade immune surveillance, enabling them to escape immune responses while still manipulating the immune landscape to their advantage. This observation is consistent with existing literature, which suggests that tumors can create an immunosuppressive environment to selectively alter T cell function [42]. Therefore, the increased presence of exhausted T cells in normal tissue compared to tumor tissue highlights a potential shift in immune regulation mechanisms between these environments. This dynamic balance of immune regulation may play a crucial role in the tumor immune response.

In the complex interplay within the TME, comprising tumor cells, immune cells, stromal cells, and other elements [43], patients were classified into high- and low-risk groups based on eight prognostic gene analyses. The low-risk group, characterized by lower tumor purity, suggested a higher presence of non-tumor entities, such as immune and stromal cells, indicating a potentially stronger immune response compared to the high-risk group. The prognostic risk model independently predicted overall survival (OS) in LIHC patients, as confirmed by subsequent analyses. This prompted an exploration of the underlying immunological mechanisms. Immune infiltration analysis showed higher expression of memory B cells, follicular helper T cells, M0 macrophages, regulatory T cells, and neutrophils in the low-risk group (P < 0.05). In contrast, the high-risk group exhibited increased expression of CD8 + T cells, resting CD4 memory T cells, M1 macrophages, both resting and activated mast cells, and eosinophils (P < 0.05). The low-risk group had higher expression of B2M, PD1, and CTLA4 (P < 0.05), whereas the high-risk group showed elevated levels of CD80, CD86, LDHA, IL12A, JAK1, LGALS9, PVR, TNFRSF18, TNFRSF9, and LGALS9 (P < 0.05). Pathway enrichment analysis indicated a dominance of ribosomal transporter metabolism in the high-risk group and increased genomic transcriptional regulation in the low-risk group. These findings elucidate the distinct tumor biological behaviors of the risk groups. Immune checkpoint analysis revealed augmented expression of PD1, CTLA4, and B2M in the low-risk group (P < 0.05), suggesting a more favorable response to immune checkpoint inhibitors. Additionally, low-risk group patients with a high tumor mutation burden (TMB) demonstrated improved OS compared to those with low TMB. Cytological experiments on hepatocellular carcinoma cell lines validated the expression of relevant prognostic genes, consistent with TCGA dataset findings. In summary, our study establishes that the constructed risk model independently predicts LIHC prognosis and offers novel biomarkers for clinical assessment and treatment planning.

This study has several limitations, primarily, it relies on data from public databases, despite utilizing multiple datasets for validation. Nevertheless, to thoroughly validate our risk model and gain a deeper understanding of the molecular mechanisms of T-cell marker genes in LIHC, it is essential to conduct corroborating in vivo and in vitro experiments. In future research, we will concentrate on elucidating the underlying molecular mechanisms of these independent prognostic factors.