Li Longhu, Li Guangyao, Zhai Wangfeng
Department of Intervention, Linfen Central Hospital, Linfen, PR China.
Department of Intervention, Linfen Central Hospital, Linfen, PR China.
Dig Liver Dis. 2025 May;57(5):611-623. doi: 10.1016/j.dld.2025.01.196. Epub 2025 Feb 3.
Hepatocellular carcinoma (HCC) is a substantial global health challenge owing to its high mortality rate and limited therapeutic options. We aimed to develop an efferocytosis-related gene signature (ER.Sig) and conduct a transcriptomic analysis to predict the prognosis and immunotherapeutic responses of patients with HCC.
Single-cell RNA sequencing data and bulk RNA sequencing data were obtained from public databases. Based on single-sample gene set enrichment analysis and Weighted Gene Co-expression Network analyses, efferocytosis-related genes (ERGs) were selected at both the single-cell and bulk transcriptome levels. A machine-learning framework employing ten different algorithms was used to develop the ER.Sig. Subsequently, a multi-omics approach (encompassing genomic analysis, single-cell transcriptomics, and bulk transcriptomics) was employed to thoroughly elucidate the prognostic signatures.
Analysis of the HCC single-cell transcriptomes revealed significant efferocytotic activity in macrophages, endothelial cells, and fibroblasts within the HCC microenvironment. We then constructed a weighted co-expression network and identified six modules, among which the brown module (168 genes) was most highly correlated with the efferocytosis score (cor = 0.84). Using the univariate Cox regression analysis, 33 prognostic ERGs were identified. Subsequently, a predictive model was constructed using 10 machine-learning algorithms, with the random survival forest model showing the highest predictive performance. The final model, ER.Sig, comprised nine genes and demonstrated robust prognostic capabilities across multiple datasets. High-risk patients exhibited greater intratumoral heterogeneity and higher TP53 mutation frequencies than did low-risk patients. Immune landscape analysis revealed that compared with high-risk patients, low-risk patients exhibited a more favorable immune environment, characterized by higher proportions of CD8+ T and B cells, tumor microenvironment score, immunophenoscore, and lower Tumor Immune Dysfunction and Exclusion scores, indicating better responses to immunotherapy. Additionally, an examination of an independent immunotherapy cohort (IMvigor210) demonstrated that low-risk patients exhibited more favorable responses to immunotherapy and improved prognoses than did their high-risk counterparts.
The developed ER.Sig effectively predicted the prognosis of patients with HCC and revealed significant differences in tumor biology and treatment responses between the risk groups.
肝细胞癌（HCC）因其高死亡率和有限的治疗选择，是一项重大的全球健康挑战。我们旨在开发一种与吞噬作用相关的基因特征（ER.Sig），并进行转录组分析，以预测HCC患者的预后和免疫治疗反应。
从公共数据库中获取单细胞RNA测序数据和批量RNA测序数据。基于单样本基因集富集分析和加权基因共表达网络分析，在单细胞和批量转录组水平上选择与吞噬作用相关的基因（ERGs）。使用包含十种不同算法的机器学习框架来开发ER.Sig。随后，采用多组学方法（包括基因组分析、单细胞转录组学和批量转录组学）来全面阐明预后特征。
对HCC单细胞转录组的分析显示，HCC微环境中的巨噬细胞、内皮细胞和成纤维细胞具有显著的吞噬活性。然后我们构建了一个加权共表达网络并鉴定出六个模块，其中棕色模块（168个基因）与吞噬作用评分的相关性最高（cor = 0.84）。使用单变量Cox回归分析，鉴定出33个预后ERG。随后，使用10种机器学习算法构建了一个预测模型，随机生存森林模型显示出最高的预测性能。最终模型ER.Sig由9个基因组成，并在多个数据集中表现出强大的预后能力。高风险患者比低风险患者表现出更大的肿瘤内异质性和更高的TP53突变频率。免疫景观分析显示，与高风险患者相比，低风险患者表现出更有利的免疫环境，其特征是CD8 + T细胞和B细胞比例更高、肿瘤微环境评分、免疫表型评分更高，以及肿瘤免疫功能障碍和排除评分更低，表明对免疫治疗的反应更好。此外，对一个独立的免疫治疗队列（IMvigor210）的检查表明，低风险患者比高风险患者对免疫治疗表现出更有利的反应和更好的预后。
所开发的ER.Sig有效地预测了HCC患者的预后，并揭示了风险组之间在肿瘤生物学和治疗反应方面的显著差异。
