Identifying potential signatures of immune cells in hepatocellular carcinoma using integrative bioinformatics approaches and machine-learning strategies.

Abstract

Hepatocellular carcinoma (HCC) is a malignant tumor regulated by the immune system. Immunotherapy using checkpoint inhibitors has shown encouraging outcomes in a subset of HCC patients. The main challenges in checkpoint immunotherapy for HCC are to expand treatment options and to broaden the beneficiary population. Therefore, the search for potential signatures of immune cells is meaningful in the development of immunotherapy for HCC. The HCC related datasets were downloaded from the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA). Differential expression analysis and functional analysis were performed first. Then support vector machine-recursive feature elimination (SVM-RFE), random forests (RF), least absolute shrinkage and selection operation (LASSO), and weighed gene co-expression network analysis (WGCNA) were employed to screen for critical genes, and receiver operating characteristic (ROC) analysis was performed to compare diagnostic performance. Subsequently, single-sample gene set enrichment analysis (ssGSEA) was used to explore the relationship between signatures and immune cells. Finally, we validated the expression of these biomarkers in human HCC samples. 531 overlapping differentially expressed genes (DEGs) were identified. Furthermore, enrichment analysis revealed pathways associated with immune activation processes, immune cell involvement and inflammatory signaling. After using multiple machine-learning strategies, extracellular matrix protein 1 (ECM1), leukemia inhibitory factor receptor (LIFR), sushi repeat containing protein X-linked (SRPX), and thromboxane A2 receptor (TBXA2R) were identified as critical signatures, and exhibited high expression in tumor-adjacent normal tissues. According to the ssGSEA results, ECM1, LIFR, SRPX and TBXA2R were all significantly associated with diverse immune cells, such as monocytes and neutrophils. Moreover, immunostaining of human HCC samples showed that these critical signatures all colocalized with CD14-positive monocytes. Our findings report the potential signatures of immune cells in HCC and confirm that they localize in monocytes of tumor-adjacent normal tissues. ECM1, LIFR, SRPX and TBXA2R could become new potential targets for predictive diagnosis, early intervention and immunotherapy of HCC in the future.