Skip to main content
Premium Trial:

Request an Annual Quote

Methylation Profiling of Cell-Free DNA Points to Tissue of Origin


NEW YORK (GenomeWeb) – Circulating cell-free DNA can be traced back to its tissue of origin through whole-genome bisulfite sequencing, according to researchers from the Chinese University of Hong Kong.

Dennis Lo, senior author of a paper describing the work and published today in the Proceedings of the National Academy of the Sciences, told GenomeWeb that the findings could have applications for cancer, noninvasive prenatal diagnostics, and transplantation.

Cell-free DNA already has a number of clinical applications, such as screening for fetal aneuploidy in pregnant women and monitoring cancer patients' response to treatment. Multiple tissues shed DNA that contributes to the total amount of circulating DNA in plasma; however, current clinical applications do not distinguish from which tissue the DNA has originated. Being able to trace the piece of DNA back to its tissue of origin could add another level of precision to existing tests and also open up new avenues of research and diagnostics, according to the researchers.

Different tissues have been found to have different methylation profiles, so the researchers decided to see whether they could use whole-genome bisulfite sequencing of cell-free DNA to create a signature that correlated with the fragment's tissue of origin.

First, the group created a "tissue map" of plasma DNA. They studied methylation profiles of 14 different tissues and selected markers that could be used to trace DNA back to the tissue. They defined two types of markers. Type I is a genomic locus that had a methylation level in one tissue that was significantly different from other tissues. A type II marker, by contrast, showed high variability in methylation densities across the panel of 14 tissues. The researchers identified 1,013 type I markers and 4,807 type II markers.

Lo said that although there are individual differences in methylation, "there is a sufficient degree of consistency" in the markers to generate what is essentially a methylation reference panel by which to trace back the DNA fragments.

In addition, plasma DNA tends to be highly fragmented, averaging about 150 bp in length. In order to maximize the possibility that a given DNA fragment would contain a methylation marker, Lo said that the team focused on CpG islands. "Such regions have a relative concentration of CpG sites that would allow a number of such sites to be present on a plasma DNA fragment," he said.

The researchers then used these markers to create an algorithm that could use methylation profiles from bisulfite sequencing to trace the DNA back to its original tissue.

The team dubbed the approach methylation deconvolution. "Each tissue has its own methylation pattern for a selected group of methylation markers," Lo explained, which is "determined by the fractional contributions of the different tissues and the methylation level of the methylation marker in the respective tissues." When analyzing the methylation levels, the methylation pattern of the mixture is determined and then compared with the profiles of the different tissues to "deduce the fractional contribution of each tissue to the mixture DNA," he explained.

First, the researchers had to figure out what tissues contributed to plasma DNA and in what proportions. Because white blood cells, the liver, and placenta during pregnancy are known to be major contributors of circulating cell-free DNA, the researchers focused on these tissues.

The team verified that the methylation signatures could distinguish DNA from each of these tissues by creating mixtures. They used plasma DNA from a healthy 40-year-old non-pregnant woman, placental DNA after the delivery of a female baby, and liver DNA obtained from noncancerous liver tissue that was adjacent to a liver tumor that had been surgically removed from a woman.

By creating various mixtures of these samples, the researchers verified that the algorithm could accurately separate them out based on methylation.

Next, they performed bisulfite sequencing on 15 pregnant women to determine the different tissues that contributed to the plasma DNA. The team found that white blood cells (neutrophils and lymphocytes) were the largest contributors, while the placenta contributed between 12 percent and 41 percent of the total circulating DNA.

Next, they examined samples from four individuals who had received liver transplants and three who had received bone marrow transplants. First, they estimated the fraction of DNA from donor and recipient by looking at SNP differences between the DNA of the donor and that of the recipient for liver and bone marrow for the respective patients. They then compared the SNP differences to their methylation deconvolution method, and found that the results were comparable.

Finally, they wanted to see whether methylation deconvolution could work in cancer patients. They performed bisulfite sequencing on 29 hepatocellular carcinoma patients and 32 control subjects without cancer. For the 32 controls and 26 of the HCC patients, genome-wide bisulfite sequencing results had previously been reported. Methylation deconvolution indicated that liver DNA contributed to about 10.7 percent of the total plasma DNA in healthy subjects, but to 24 percent in the individuals with liver cancer. In addition, the researchers had access to tumor tissue from 14 of the individuals with HCC and found that tumor DNA concentration correlates with the percentage of liver DNA circulating in cancer patients' plasma.

After showing that the principal of methylation deconvolution could tell the proportion of plasma DNA originating from placenta, liver, and white blood cells, the researchers wanted to see whether specific copy number alterations could be linked to the specific tissue of origin. For instance, noninvasive prenatal tests screen for fetal chromosomal abnormalities, but sometimes an aberrant result on a test turns out to be not of fetal origin, but actually from a maternal malignancy. So a method that could tell the tissue of origin of a copy number alteration could help distinguish between these cases.

The researchers used data from bisulfite sequencing that had previously been performed on plasma DNA from five pregnant women carrying fetuses with trisomy 21. The team hypothesized that when evaluating methylation markers on the genomic locus associated with the CNV, there would be an increase of markers associated with the tissue of origin for copy number gains and a decrease of the markers associated with the tissue of origin for a loss, when compared to the reference.

When they tested this on the five samples positive for fetal trisomy 21, they found that in all cases, there were a greater-than-expected number of methylation markers associated with placental DNA at chromosome 21, but not for any other chromosome.

The researchers then tested this hypothesis on liver cancer patients. Seven of the HCC patients contained a CNV of at least 30 mb in size, and the researchers found that, after performing methylation deconvolution and then evaluating the regions with CNVs, those fragments could be traced back as being from liver tissue.

Lo said that the study is a proof of concept, but he sees the potential for future clinical applications, particularly in the areas of NIPT, oncology, and transplantation monitoring.

For instance, he said, the method could be used in "cancer detection from plasma, or in the incidental detection of cancer over the course of NIPT." Currently, he said, there have been a few examples of women that have received NIPT only to find out that they have cancer. In these cases, the result leads to further diagnostic tests to try and identify the location of the tumor, Lo said. But, "the methylation deconvolution approach would allow us to point towards likely tissues that we should focus on," rather than going immediately to a whole-body scan.

Lo estimated that the whole-genome bisulfite sequencing approach would currently cost around $2,500 per sample, which he acknowledged would be too expensive for current NIPTs, particularly since identifying maternal malignancies in this context is still so rare. However, he said that in cases where a potential cancer alteration is identified in the course of an NIPT, only then would bisulfite sequencing by used. For instance, he said, a recent paper found that in cases where a NIPT identifies more than one chromosomal aberration, 18 percent turn out to be cancer, rather than a fetus with multiple trisomies. "I think that such women would be a good cohort to test the methylation deconvolution approach," Lo said.

The group next wants to explore the method for multiple cancer types and "further refine our marker list based on additional tissue methylome data that are generated by various groups," he added.