Noninvasive prognostication of hepatocellular carcinoma based on cell-free DNA methylation

Abstract

Background: The current noninvasive prognostic evaluation methods for hepatocellular carcinoma (HCC), which are largely reliant on radiographic imaging features and serum biomarkers such as alpha-fetoprotein (AFP), have limited effectiveness in discriminating patient outcomes. Identification of new prognostic biomarkers is a critical unmet need to improve treatment decision-making. Epigenetic changes in cell-free DNA (cfDNA) have shown promise in early cancer diagnosis and prognosis. Thus, we aim to evaluate the potential of cfDNA methylation as a noninvasive predictor for prognostication in patients with active, radiographically viable HCC.

Methods: Using Illumina HumanMethylation450 array data of 377 HCC tumors and 50 adjacent normal tissues obtained from The Cancer Genome Atlas (TCGA), we identified 158 HCC-related DNA methylation markers associated with overall survival (OS). This signature was further validated in 29 HCC tumor tissue samples. Subsequently, we applied the signature to an independent cohort of 52 patients with plasma cfDNA samples by calculating the cfDNA methylation-based risk score (methRisk) via random survival forest models with 10-fold cross-validation for the prognostication of OS.

Results: The cfDNA-based methRisk showed strong discriminatory power when evaluated as a single predictor for OS (3-year AUC = 0.81, 95% CI: 0.68-0.94). Integrating the methRisk with existing risk indices like Barcelona clinic liver cancer (BCLC) staging significantly improved the noninvasive prognostic assessments for OS (3-year AUC = 0.91, 95% CI: 0.80-1), and methRisk remained an independent predictor of survival in the multivariate Cox model (P = 0.007).

Conclusions: Our study serves as a pilot study demonstrating that cfDNA methylation biomarkers assessed from a peripheral blood draw can stratify HCC patients into clinically meaningful risk groups. These findings indicate that cfDNA methylation is a promising noninvasive prognostic biomarker for HCC, providing a proof-of-concept for its potential clinical utility and laying the groundwork for broader applications.

Fig 1. Schematic diagram of the study design.

We utilized the TCGA dataset in a two-step feature selection process to identify 158 methylation markers that are associated with HCC and survival. We validated these markers’ prognostic value in our HCC tumor tissue samples. We then applied random survival forest with 10-fold cross-validation to predict the cfDNA methylation-based overall survival risk score (methRisk) for plasma samples of an independent patient cohort. Subsequent functional analysis and various metrics were used to evaluate the performance of the methRisk. Figure created with BioRender.com.

Fig 2. Functional analysis of DNA methylation markers and cfDNA-based methRisk performance.

(A) Composition of markers in terms of genome annotation. (B) KEGG pathway enrichment analysis, and GO enrichment analysis of biological pathways associated with the methylation marker-related genes. The bar charts show the top 10 enriched terms ranked by p-value. (C) Time-dependent AUCs for individual features predicting 3-year survival in the patient cohort with plasma samples. Error bars represent the 95% confidence intervals. The sample size (n) is annotated to the right of the error bar.

Fig 3. Box plots illustrating the methRisk distribution among HCC patient subgroups in the patient cohort with plasma samples.

Patients were stratified based on prior AFP, largest tumor size, tumor number, MELD score, BCLC, and transplant criteria. P-values from the one-sided Mann-Whitney-Wilcoxon test comparing methRisk between subgroups are indicated above (* P < 0.05). The median methRisk value of each patient subgroup is denoted by a red line.

Fig 4. Kaplan-Meier survival curves for overall survival among HCC patient subgroups in the patient cohort with plasma samples.

Patients were stratified based on median values of prior AFP, largest tumor size, MELD score, and methRisk. The low-risk group (blue) and high-risk group (red) were compared using a two-sided log-rank test, with the associated p-value indicated. Hazard ratios were calculated via univariate Cox proportional-hazard models.

Fig 5. Forest plot displaying the hazard ratios of each feature in univariate analysis in the patient cohort with plasma samples.

Patients were stratified into subgroups based on individual features, with the respective group sizes presented. Hazard ratios were estimated using univariate Cox proportional-hazard models and error bars represent the 95% confidence intervals. The p-value is based on the Wald test to assess the significance of the coefficient of a predictor (* P < 0.05, ** P < 0.01).

Fig 6. Enhancing noninvasive prognostic prediction with methRisk.

(A) The performances of features for OS prediction were assessed in the patient cohort with plasma samples. Error bars represent the 95% confidence intervals. The sample size (n) is annotated above the error bar. (B) A composite nomogram that integrates BCLC and methRisk for predicting 2-year, 3-year, and 4-year survival. To obtain the predicted survival probability, sum up the points corresponding to each feature and trace a vertical line from the total points to the survival rate scale. (C) The Kaplan-Meier survival curve for CCFI between HCC patients stratified by the median value of methRisk. The low-risk group (blue) and high-risk group (red) were compared using a two-sided log-rank test, with the associated p-value indicated. The HR was calculated via a univariate Cox proportional-hazard model.