The Concept Cancer cells are characterized by widespread loss of DNA methylation (genome-wide hypomethylation) relative to normal cells. This study explored whether this epigenetic change could be detected in cell-free DNA shed by tumors into blood plasma.
The Approach Shotgun massively parallel bisulfite sequencing of plasma DNA was used to detect genome-wide hypomethylation in patients with hepatocellular carcinoma and other cancers, using a cost-effective and genome-wide approach.
Dual Biomarker A key innovation is that the same bisulfite sequencing data can simultaneously detect both DNA hypomethylation and tumor-associated copy number aberrations (CNAs) in plasma, doubling the information from a single assay.
Patient Cohorts Twenty-six HCC patients (25 BCLC stage A, 1 stage B) and 32 healthy subjects were the primary study group. An additional 20 non-HCC cancer patients (breast, lung, nasopharyngeal, sarcoma, neuroendocrine) and 8 HBV/cirrhosis patients were also analyzed.
Sequencing Depth Full-depth sequencing used approximately 93 million aligned reads per case (one lane of Illumina HiSeq 2000). Downsampling to 10 million reads was tested to assess cost reduction feasibility.
Methylation Analysis The genome was divided into 1 Mb bins. Methylation density in each bin was compared to healthy controls using z-score analysis. Bins with methylation density 3 standard deviations below control mean were called hypomethylated.
Hypomethylation Signal HCC patients showed a median of 34.1% of genome bins with significant hypomethylation, compared to 0% in healthy controls. The AUC for HCC detection by hypomethylation analysis was 0.93 (95% CI 0.87-1.00).
Sensitivity and Specificity At the optimal cutoff of 1.1% hypomethylated bins, hypomethylation analysis achieved 81% sensitivity and 94% specificity for detecting HCC. At 10 million reads sequencing depth, sensitivity was 68% with maintained 94% specificity.
CNA Analysis Using the same data for copy number analysis achieved 81% sensitivity and 88% specificity with 93 million reads. Combining hypomethylation and CNA using an 'OR' algorithm boosted sensitivity to 92% with 88% specificity.
Broad Cancer Applicability The approach was applied to 20 non-HCC cancer patients including breast, lung, nasopharyngeal, smooth muscle sarcoma, and neuroendocrine tumor. Across all 38 nonmetastatic cases, the combined 'OR' algorithm achieved 87% sensitivity at 88% specificity.
Metastatic vs. Nonmetastatic Patients with metastatic disease uniformly showed the highest proportions of hypomethylated bins, consistent with larger tumor burden releasing more ctDNA into the blood. This validates that the signal reflects the tumor fraction in plasma.
Low Sequencing Depth Feasibility For hypomethylation analysis, performance at 10 million reads was nearly identical to full sequencing depth (AUC 0.96 vs. 0.93), making this approach potentially cost-effective for population-scale screening.
Serial Monitoring In two HCC patients with pre-operative and post-operative plasma samples, hypomethylation signals declined sharply after surgical resection, consistent with clearance of tumor-derived DNA from circulation.
Residual Disease Detection In patient TBR34, hypomethylation remained elevated at 2 months post-resection, foreshadowing subsequent discovery of multifocal intrahepatic metastases and lung metastases leading to death at 8 months.
Sustained Remission In patient TBR36, hypomethylation became undetectable within 3 months of resection and remained negative at 12 months, correlating with continued clinical remission at 20 months post-surgery.
Larger Validation Needed The sample sizes in this study are modest. Prospective validation in large cohorts including patients at various cancer stages and high-risk individuals (e.g., cirrhosis without HCC) is essential before clinical adoption.
False Positive Concern One cirrhosis patient without HCC had a positive hypomethylation result. Long-term follow-up of such individuals will determine whether positive signals in high-risk non-cancer individuals represent early disease or truly false positives.
Single-Molecule Sequencing Emerging single-molecule sequencing platforms can detect methylation directly without bisulfite conversion, potentially simplifying the workflow. The integration of hydroxymethylation and other epigenetic marks may further improve cancer detection specificity.