Multi-Omics Decodes Metastatic Cellular States at the HCC Invasion Front

Hepatocellular carcinoma (HCC) metastasis is driven by complex transcriptomic alterations and profound cellular heterogeneity. To elucidate the underlying mechanisms, we integrated bulk RNA sequencing (TCGA-LIHC; 379 tumor and 59 normal samples), single-cell RNA sequencing (scRNA-seq, GSE149614; 71,915 cells from 10 patients), and spatial transcriptomics (ST, HRA000437; 84,823 spots across 21 tissue slices from 7 patients). First, using LASSO-penalized Cox regression on bulk RNA-seq data, we constructed a robust 7-gene metastasis-associated prognostic model (including FAM180A and FCN2, p < 0.001). Projecting this macroscopic signature onto our single-cell atlas revealed that these prognostic programs are highly enriched in specific malignant hepatocyte subpopulations associated with portal vein tumor thrombus (PVTT) and lymph node metastasis (LNM). Pseudotime trajectory analysis confirmed a dynamic transcriptomic transition of these hepatocytes toward a highly invasive state during tumor progression. Crucially, advancing beyond traditional multi-omics HCC studies, we mapped these high-risk cellular states directly onto ST data, demonstrating their preferential localization at the tumor invasion front. Furthermore, spatially resolved differential expression analysis identified a novel panel of leading-edge (LE) specific biomarkers-notably MARCO (log2FC = 4.987, p = 1.22E-94) and TTC36 (log2FC = 4.139, p = 5.57E-163)-that are stably upregulated at the invasion front across multiple patients. Together, this multidimensional approach delineates the transcriptomic evolution of prognostic cellular states and provides a space-specific blueprint for HCC metastasis, offering novel localized targets for intercepting tumor invasion.