Noninvasive detection of cancer-associated genome-wide hypomethylation and copy number aberrations by plasma DNA bisulfite sequencing

Abstract

We explored the detection of genome-wide hypomethylation in plasma using shotgun massively parallel bisulfite sequencing as a marker for cancer. Tumor-associated copy number aberrations (CNAs) could also be observed from the bisulfite DNA sequencing data. Hypomethylation and CNAs were detected in the plasma DNA of patients with hepatocellular carcinoma, breast cancer, lung cancer, nasopharyngeal cancer, smooth muscle sarcoma, and neuroendocrine tumor. For the detection of nonmetastatic cancer cases, plasma hypomethylation gave a sensitivity and specificity of 74% and 94%, respectively, when a mean of 93 million reads per case were obtained. Reducing the sequencing depth to 10 million reads per case was found to have no adverse effect on the sensitivity and specificity for cancer detection, giving respective figures of 68% and 94%. This characteristic thus indicates that analysis of plasma hypomethylation by this sequencing-based method may be a relatively cost-effective approach for cancer detection. We also demonstrated that plasma hypomethylation had utility for monitoring hepatocellular carcinoma patients following tumor resection and for detecting residual disease. Plasma hypomethylation can be combined with plasma CNA analysis for further enhancement of the detection sensitivity or specificity using different diagnostic algorithms. Using the detection of at least one type of aberration to define an abnormality, a sensitivity of 87% could be achieved with a specificity of 88%. These developments have thus expanded the applications of plasma DNA analysis for cancer detection and monitoring.

Keywords: epigenetics; epigenomics; global hypomethylation; next-generation sequencing; tumor markers.

Fig. 1.

Variations of plasma methylation density (MD) of the 32 healthy subjects for each 1 Mb according to the genomic coordinates. The numbers within each box represent the chromosome number.

Fig. 2.

Methylation analyses for a representative HCC patient (HOT215) and a healthy control subject. For the HCC patient, the methylation densities of each 1-Mb bin for the tumor tissue (purple) and buffy coat (black) are shown in the inner ring. For the MD in tissue, the range shown is from 0% (innermost) to 100% (outermost) and the distance between two lines is 10%. The methylation z scores of the patient's plasma are shown in the middle ring. For the healthy control subject, the methylation z scores of the plasma are shown in the outer ring. For the plasma analysis, the red and gray dots represent bins with and without hypomethylation, respectively.

Fig. 3.

Percentage of bins showing hypomethylation in HCC patients and chronic hepatitis B virus (HBV) carriers with cirrhosis (A). ROC curves for HCC patient detection using hypomethylation (B) and CNA (C) analyses. The P values shown are for the comparisons between using all sequenced reads from one lane and using 10 million reads.

Fig. 4.

Serial analysis for plasma methylation and CNA for cases TBR34 (A) and TBR36 (B). For methylation analysis, the innermost ring shows the MD of the buffy coat (black) and tumor tissues (purple). For CNA analysis, the innermost ring shows the CNA detected in the tumor tissues. For both types of analyses, from the second innermost ring outward, the plasma results for different time points are shown. (A) For TBR34, the plasma samples were taken before surgery, at 3 d and 2 mo after tumor resection, respectively. (B) For TBR36, the plasma samples were taken before surgery, at 3 d, 3 mo, 6 mo, and 12 mo after tumor resection, respectively.

Fig. 5.

Plasma hypomethylation and CNA analyses for three representative patients suffering from breast, lung, and nasopharyngeal cancers. The inner ring and outer ring show the CNA and methylation z scores, respectively. Each dot represents a 1-Mb bin. For CNA analysis, the green, red, and gray dots represent bins with chromosome gain, loss, and normal chromosome dosage, respectively. For methylation analysis, the red and gray dots represent bins with and without hypomethylation, respectively. The distance between two parallel lines represents a z-score difference of 5.

Fig. 6.

Diagnostic performance for hypomethylation and CNA analyses for patients with metastatic and nonmetastatic cancers. (A) Percentage of bins showing hypomethylation. The red circles and blue rhombuses represent cancer patients with HCC and non-HCC, respectively. (B) ROC curves for the detection of all cases with nonmetastatic cancers using hypomethylation and CNA analyses. The P values were for the comparisons between using all sequenced reads of one lane (i.e., a mean of 93 million reads) and using 10 million reads.