Plasma proteomic signature for preoperative prediction of microvascular invasion in HCC.

Abstract

Background & aims: Microvascular invasion (MVI) is a major determinant of poor prognosis in hepatocellular carcinoma (HCC). However, reliable non-invasive biomarkers for the preoperative evaluation and diagnosis of MVI are urgently needed in clinical practice.

Methods: Plasma samples were collected from 160 patients with HCC (80 MVI-positive and 80 MVI-negative) from four medical centers. Plasma proteomic profiling was obtained using data-independent acquisition mass spectrometry. Principal component analysis and differential protein abundance analysis were used to assess the proteomic changes between the two groups of patients. Protein biomarker candidates were further quantitatively validated by ELISA.

Results: Proteomic analysis of 50 patients with HCC (25 MVI-positive and 25 MVI-negative) identified three plasma protein biomarkers (TALDO1, PDIA3, and PGK1) that are significantly upregulated in MVI-positive patients (FDR-adjusted p <0.05) and were subsequently cross-validated by ELISA. A machine learning-based Plasma pRotein MVI risk Model (PRIM) was developed for the preoperative prediction of MVI. PRIM demonstrated excellent discriminatory ability, with areas under the receiver operating characteristic curve values ranging from 0.78 to 0.99 across three independent cohorts. Single-cell RNA sequencing of five HCC tumors provided a cell type-resolved atlas of biomarker expression, showing their predominant presence in malignant cells and macrophages within the MVI-positive tumor microenvironment compared with MVI-negative tumors.

Conclusions: This study provides a comprehensive analysis of the plasma proteomic landscape in HCC and presents a promising blood-based tool for preoperative MVI risk stratification.

Impact and implications: This study highlights the transformative potential of plasma proteomic profiling in improving the preoperative prediction of microvascular invasion in hepatocellular carcinoma (HCC). By integrating data-independent acquisition mass spectrometry and machine learning, we identified three plasma protein biomarkers (TALDO1, PDIA3, and PGK1) and developed the Plasma pRotein MVI risk Model (PRIM), which demonstrated robust diagnostic accuracy across multicenter validation cohorts. These findings pave the way for preoperative risk stratification and personalized therapeutic strategies in HCC management.