Construction of a Prognostic and Diagnostic Gene Signature Based on the Characteristics of Cancer Stem‐Like Cells for the Treatment Guidance of Hepatocellular Carcinoma

BACKGROUND: Hepatocellular carcinoma (HCC) is a highly aggressive cancer with significant intratumoral heterogeneity, likely contributed by the differentiation of cancer stem-like cells (CSCs) into multiple cancer cell lineages, resulting in therapeutic resistance and increased patient mortality. However, there is limited evidence from patient samples regarding the critical role of CSCs in treatment efficacy and patient outcomes. METHODS: In this study, through the analysis of single-cell transcriptomic data of HCC patient samples, we identify a rare population of cells with cancer stem-like features. Next, we constructed a 12-gene signature that accurately predicts patient survival. Furthermore, based on the expression levels of the 12-gene signature of CSCs, a CAncer stem-like cell Risk Score (CARS) is computed. RESULTS: The results show that patients with a higher CARS exhibit increased resistance to sorafenib but greater sensitivity to immune checkpoint inhibitors (ICI). The identification of CSCs in the single-cell transcriptome also provides a unique opportunity to study the microenvironmental characteristics of CSCs in HCC. HCC samples with higher CARS show a substantial infiltration of B cells, and the glycoprotein signaling of GRN and PSAP is specifically associated with the crosstalk between B cells and CSCs, rather than other tumor cells. CONCLUSION: These findings develop CARS of potential prognostic and predictive value in HCC management.