NEW YORK — Childhood bone cancers may be able to be detected through liquid biopsies based on their epigenetic patterns, a new study has found.
Liquid biopsies in which cell-free DNA sloughed off from tumor cells have increasingly been used as diagnostic and prognostic tools in oncology and could help personalize and monitor patients' treatments. Many of the analysis approaches, however, rely on the detection of known tumor-linked genetic alterations. As pediatric cancers typically have a low rate of recurrent genomic changes, a team led by researchers from St. Anna Children's Cancer Research Institute in Vienna sought a different way of identifying tumor-derived cell-free DNA.
As they reported in Nature Communications on Friday, the researchers instead focused on the size of the cell-free DNA fragments they uncovered within their cohort of pediatric patients with Ewing sarcoma, a type of bone tumor. The fragment sizes, they noted, reflect epigenetic patterns of the cells from which they originated — including tumor cells.
"We previously identified unique epigenetic signatures of Ewing sarcoma. We reasoned that these characteristic epigenetic signatures should be preserved in the fragmentation patterns of tumor-derived DNA circulating in the blood," co-senior author Eleni Tomazou from St. Anna CCRI said in a statement. "This would provide us with a much-needed marker for early diagnosis and tumor classification using the liquid biopsy concept."
The researchers further developed an algorithm to detect tumor-derived DNA based on their chromatin signatures, which they said could be used as both a diagnostic and prognostic tool.
For their study, Tomazou and her colleagues sequenced 263 cell-free DNA samples from 95 Ewing sarcoma patients, 31 patients with other pediatric sarcomas, and 22 healthy controls. This set of samples included ones from the same patients at different disease stages.
After confirming that they could detect tumor-derived DNA in their cell-free DNA samples, the researchers noted that the tumor-derived DNA tended to be shorter than the cell-free DNA found among healthy controls. These fragments were often about 167 base pairs in size, which corresponds to the length of DNA that wraps around a nucleosome plus linker DNA.
These shorter tumor-derived DNA fragments were additionally found among patients for whom there were no detectable genetic changes. This suggested to the researchers that fragmentation patterns could be used to identify tumor-derived DNA independent of genetic alterations.
The researchers further developed a tool they dubbed LIQUORICE — for liquid biopsy regions-of-interest coverage estimation — that overlays these genome-wide cell-free DNA fragment profiles atop a predetermined set of genomic regions where epigenetic changes often occur in the cancer type. In that way, it develops a consensus signature of fragment coverage in those regions.
They additionally developed machine-learning classifiers to not only distinguish individuals with cancer from controls but also between different types of pediatric sarcoma, without relying on the recurrent genetic changes. When they tested their classifiers, the researchers found they were highly sensitive. "By feeding these machine-learning algorithms with our extensive whole-genome sequencing data of tumor-derived DNA in the bloodstream, the analysis becomes highly sensitive and in many instances outperforms conventional genetic analyses," Tomazou said.
She and her colleagues noted, though, that their findings need to be validated in a large, prospective study. Still, they said that their assay has the potential to be used as a minimally invasive diagnostic as well as a prognostic marker and could also be used to gauge treatment response and tailor treatments based on the response measured.