DNA methylation biomarker panels for differentiating various liver adenocarcinomas, including hepatocellular carcinoma, cholangiocarcinoma, colorectal liver metastases and pancreatic adenocarcinoma liver metastases

Abstract

Background: DNA methylation biomarkers are one of the most promising tools for the diagnosis and differentiation of adenocarcinomas of the liver, which are among the most common malignancies worldwide. Their differentiation is important because of the different prognoses and treatment options. This study aimed to validate previously identified DNA methylation biomarkers that successfully differentiate between liver adenocarcinomas, including the two most common primary liver cancers, hepatocellular carcinoma (HCC) and cholangiocarcinoma (CCA), as well as two common metastatic liver cancers, colorectal liver metastases (CRLM) and pancreatic ductal adenocarcinoma liver metastases (PCLM), and translate them to the methylation-sensitive high-resolution melting (MS-HRM) and digital PCR (dPCR) platforms.

Methods: Our study included a cohort of 149 formalin-fixed, paraffin-embedded tissue samples, including 19 CRLMs, 10 PCLMs, 15 HCCs, 15 CCAs, 15 colorectal adenocarcinomas (CRCs), 15 pancreatic ductal adenocarcinomas (PDACs) and their paired normal tissue samples. The methylation status of the samples was experimentally determined by MS-HRM and methylation-specific dPCR. Previously determined methylation threshold were adjusted according to dPCR data and applied to the same DNA methylation array datasets (provided by The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO)) used to originally identify the biomarkers for the included cancer types and additional CRLM projects. The sensitivities, specificities and diagnostic accuracies of the panels for individual cancer types were calculated.

Results: In the dPCR experiment, the DNA methylation panels identified HCC, CCA, CRC, PDAC, CRLM and PCLM with sensitivities of 100%, 66.7%, 100%, 86.7%, 94.7% and 80%, respectively. The panels differentiate between HCC, CCA, CRLM, PCLM and healthy liver tissue with specificities of 100%, 100%, 97.1% and 94.9% and with diagnostic accuracies of 100%, 94%, 97% and 93%, respectively. Reevaluation of the same bioinformatic data with new additional CRLM projects demonstrated that the lower dPCR methylation threshold still effectively differentiates between the included cancer types. The bioinformatic data achieved sensitivities for HCC, CCA, CRC, PDAC, CRLM and PCLM of 88%, 64%, 97.4%, 75.5%, 80% and 84.6%, respectively. Specificities between HCC, CCA, CRLM, PCLM and healthy liver tissue were 98%, 93%, 86.6% and 98.2% and the diagnostic accuracies were 94%, 91%, 86% and 98%, respectively. Moreover, we confirmed that the methylation of the investigated promoters is preserved from primary CRC and PDAC to their liver metastases.

Conclusions: The cancer-specific methylation biomarker panels exhibit high sensitivities, specificities and diagnostic accuracies and enable differentiation between primary and metastatic adenocarcinomas of the liver using methylation-specific dPCR. High concordance was achieved between MS-HRM, dPCR and bioinformatic data, demonstrating the successful translation of bioinformatically identified methylation biomarkers from the Illumina Infinium HumanMethylation450 BeadChip (HM450) and lllumina MethylationEPIC BeadChip (EPIC) platforms to the simpler MS-HRM and dPCR platforms.

Keywords: Cholangiocarcinoma; DNA methylation; Diagnostic biomarkers; Digital PCR; Hepatocellular carcinoma; Liver metastases; MS-HRM; Metastatic colorectal cancer; Metastatic pancreatic ductal adenocarcinoma; Primary liver cancer.

Fig. 1

Study workflow. DNA methylation biomarker candidates were selected from our published work. After sample collection, the formalin-fixed, paraffin-embedded tissue blocks were punched. DNA was isolated, quantified and subjected to bisulfite conversion. Screening of candidate DNA methylation biomarkers and initial assessment of methylation status was performed using MS-HRM. The best biomarkers were selected and grouped into panels. The panels were tested on a subset of the same samples using dPCR. Statistical analysis of the experimental results was performed. The lower dPCR methylation thresholds were assessed using publicly available DNA methylation data from the TCGA and GEO databases. The selected promoter regions were annotated and their biological function in cancer was investigated. Created with BioRender.com

Fig. 2

Graphical representation of the dPCR results. Boxplots showing the distribution of the highest beta values of all included samples for the probes of the panels. The dashed line in each boxplot presents methylation threshold for the corresponding biomarker panel. A dPCR results for HCC panel showing the distribution of methylation percentage across HCC, CCA, CRLM, PCLM and healthy liver tissue samples. B dPCR results for CCA panel showing the distribution of methylation percentage across HCC, CCA, CRLM, PCLM and healthy liver tissue samples. C dPCR results for CRC panel showing the distribution of methylation percentage across HCC, CCA, CRC, CRLM, PCLM and healthy liver tissue samples. D dPCR results for PDAC panel showing the distribution of methylation percentage across HCC, CCA, CRLM, PDAC, PCLM and healthy liver tissue samples

Fig. 3

The receiver operating characteristic (ROC) and the area under the curve (AUC) representing the overall diagnostic accuracy of the developed diagnostic panels for dPCR results. In addition to the area under the curve, a 95% confidence interval and a p value were calculated. The analysis included the methylation percentage data of adenocarcinomas (HCC, CCA, CRLM and PCLM) and healthy liver tissue (paired NATs of HCC and CCA) included in the dPCR experiment. A HCC panel ROC curve. B CCA panel ROC curve. C CRC panel ROC curve. D PDAC panel ROC curve. Abbreviations: AUC, area under the curve; CI, confidence interval