A Three-Gene Prognostic Signature Driven by an ER Stress-Associated ceRNA Network: Integrating Single-Cell...

The progression and immune escape of HCC are closely regulated by endoplasmic reticulum stress (ERS). However, the associated ceRNA regulatory networks and their prognostic value remain to be systematically elucidated. Here, we sought to establish a prognostic signature derived from an ERS-associated ceRNA network and to investigate its relationship with the tumor immune microenvironment. We integrated TCGA-LIHC transcriptomic data with the MSigDB ERS gene set to identify ERS-associated differentially expressed genes and construct a ceRNA regulatory network. Using a forward search strategy with 10-fold cross-validation, we screened candidate genes to select the optimal prognostic combination and constructed a multigene Cox regression signature. External validation was performed in the independent microarray cohort GSE14520. By integrating single-cell transcriptomics, CIBERSORT, ESTIMATE, TIDE, and drug sensitivity analyses, we revealed immune microenvironment characteristics associated with this signature. Based on the ceRNA network's eight core ERS mRNAs, an optimal three-gene signature comprising STC2, CKS1B, and PSAT1 was selected via forward search. The signature demonstrated robust prognostic discrimination in the TCGA training cohort (C-index = 0.653) and was independently corroborated in the external GSE14520 dataset (C-index = 0.584, log-rank p = 0.008). The signature was confirmed as an independent prognostic indicator by multivariable Cox regression. Functional enrichment analysis demonstrated a marked accumulation of cell-cycle-related pathways in the high-risk group, notably DNA replication and the spindle assembly checkpoint. Single-cell transcriptomic profiling showed that STC2 and CKS1B were predominantly expressed by tumor epithelial cells, whereas CCL2 and ATF3 were mainly detected in macrophages and fibroblasts. Drug sensitivity analysis indicated that the high-risk group was more sensitive to drugs such as docetaxel and AZD5582, consistent with the upregulation of proliferation pathways in this group; in the low-risk group, VE-822 exhibited selective sensitivity. This study established a three-gene prognostic signature based on the ERS-associated ceRNA network. The signature demonstrated robust prognostic stratification capabilities in cross-platform validation and revealed molecular characteristics centered on uncontrolled cell-cycle progression, as well as an immunosuppressive microenvironment, in the high-risk group, providing an exploratory tool for prognostic assessment and treatment strategy selection in hepatocellular carcinoma (HCC).