NEW YORK – A multi-lab group led by researchers at the Translational Genomic Research Institute (TGen) has built a liquid biopsy framework that analyzes fragmentation patterns in cell-free DNA (cfDNA) linked to pediatric and pancreatic cancers.
The team's non-invasive method, which analyzes fragment ends within recurrently protected regions (RPR) in cfDNA in a patient's urine sample, produced an area under the curve (AUC) of 0.89 in in terms of differentiating cancer patients from healthy subjects.
The project also involved researchers from Phoenix Children's Hospital, Baylor Scott and White Research Institute, and City of Hope.
Muhammad Murtaza, who now serves as an associate professor at the University of Wisconsin-Madison's Center for Human Genomics, explained that his team at TGen originally debated whether urine could be as adequate an analyte as blood for cfDNA analysis.
In a proof-of-principle study published last month in Science Translational Medicine, Murtaza and his colleagues used whole-genome sequencing (WGS) on 30 urine and 15 plasma samples from healthy patients to determine cfDNA fragment size distribution.
While the researchers observed an average fragment size of 167 bp in plasma samples, they found that fragment sizes in urine were on average 80 to 81 base pairs in length.
Murtaza therefore speculated that urine cfDNA was protected from degradation due to some level of immediate protection. He also noted that fragmentation patterns in plasma cfDNA are consistent with nucleosome positioning and DNA-protein interactions from contributing cells.
The team therefore decided to observe different regions of the genome that it knew were conserved and see what distribution coverage looked like in urine as opposed to plasma.
"We found that coverage was very consistent between the two samples in regions of the genome where the nucleosome positions are conserved across all cells," Murtaza said. "While the peaks are generally narrower in urine, they overlap with the centers of peaks in plasma samples."
Murtaza's team then widened the analysis to the entire genome to potentially identify and compare RPRs conserved across plasma and urine samples. The researchers found that RPR peaks overlap between different samples and that they also existed independently of plasma in urine-based RPR maps.
Murtaza and his colleagues then speculated whether cell-free fragmentation in urine and plasma retained any of the DNA's features from their cells of origin. The group thus analyzed open and closed chromatin regions in a lymphoblast cell line in plasma samples.
"When you look at plasma, on average, you find that the fragment size in closed chromatin regions are slightly longer than the fragment size in open chromatin regions," Murtaza said. "We found a similar trend for fragment size in urine samples."
Murtaza's team then tried to identify local differences in fragmentation across the entire genome. Using different cell lines, tissues, and DNA hypersensitivity sites, the group investigated correlations in plasma and urine samples.
The researchers saw the highest correlation between lymphoid and myeloid cells for plasma, compared to epithelial cells and urine, which suggested that epithelial cells contribute more cfDNA in urine compared to plasma.
"This gives us reassurance that the fragmentation features we are looking at are related to the cells that are contributing to them," Murtaza said.
Murtaza and his colleagues then aimed to see if RPRs were protected by degradation in plasma and urine samples. Importantly, they found that RPR peaks and their centers were largely spared from degradation in plasma, while the fragmentation ends fall on the periphery across the genome.
"You see similar things in the urine, although the centers' protection might be more transient because there's a greater percentage of fragments that break in the RPR," Murtaza said. "It seemed that DNA-protein interactions give rise to RPRs, based on where the the DNA is bound to the protein and their positioning in the cell."
Murtaza's group hypothesized that because cancer patients have a different tissue type, the cell contributes more DNA into urine in these patients. He pointed out that cancer cells have differences in where the DNA and proteins interact.
The group evaluated urine samples from 8 healthy and 22 stages I-IV cancer patients (10 pediatric and 12 pancreatic cancer), noting that the fraction of aberrant fragments (FAF) was higher in urine cancer patients compared to their controls. Using thresholds for FAFs and fragment end motifs, the researchers saw that the method produced an AUC of 0.89 when distinguishing between healthy controls and cancer patients.
"This gave us the remarkable ability to distinguish between cancer and healthy patients, which was improved even further when we included a multi-dynamic analysis of nucleotide sequences [that] we observed," Murtaza said.
To ensure the elevations in fragment-based RPRs were linked to the patients' tumor and not from other physiological responses, Murtaza's team compared FAFs in genomic regions with copy number gains and losses in tumor DNA. The researchers saw that the FAFs in urine increased in regions where the patients' tumor DNA carried a copy number gain, even though the copy number changes were below the limit of detection in urine circulating tumor DNA (ctDNA).
"This is consistent with the fact that the patient had earlier-stage disease, and we didn't expect high tumor fraction in plasma or urine," Murtaza noted. "But we do see that the fraction of aberrant fragments is elevated in copy number of gained regions compared to that in lost regions."
Murtaza then wanted to see if urine samples could be collected outside the lab and still behave similarly in terms of degradation patterns. The researchers selected five healthy volunteers who collected an initial voided urine sample at home and a subsequent sample at the lab later that day.
After dividing each sample into five aliquots and adding preservative material to the first aliquot, the team added preservative material to subsequent aliquots at 30 minutes, 60 minutes, 120 minutes, and 240 minutes.
Murtaza and his colleagues found that cfDNA yields and fragment size distribution were unaffected by collection of urine samples in cups with pre-added preservative material. In addition, the urine samples lasted at least 45 minutes before degradation during sample processing after collection without preservative material.
However, Murtaza acknowledged that the study had certain limitations, including a small sample size and lack of longitudinal data. The team also did not record the sources of biological and preanalytical variation during urine collection for most of the samples.
While the method may help measure RPRs in cancer patients, the study authors also acknowledged that it may not reliably capture nucleosome positions in contributing cell types, including epithelial and lymphoblasts.
"We found that fragmentation analysis is feasible in urine DNA and only requires shallow WGS, unlike deeper sequencing needed to investigate individual mutations," Murtaza said. "This limits the volume of sample needed, keeps the cost of the assay low, and provides a fast approach for performing larger studies and eventually routine testing."
Murtaza and his colleagues are now investigating how far they can push the method to see how much sequencing is needed to measure aberrant fragments in plasma samples. While the group will validate the method in larger pediatric and pancreatic cancer patient cohorts, Murtaza noted the method is not limited to one or two tumor types.
"A larger prospective study is needed with appropriate power-analysis to develop a DNA test for an accurate assessment of whether the patient is suffering from cancer or not," he added.
Commercial potential
Murtaza and his colleagues have filed a patent with the US Patent and Trademark Organization for the method that analyzes fragmentation pattern in urine cfDNA.
"The hope is, with further validation, that we'll use urine aberrant fragmentation in the context of early cancer detection," Murtaza said. "Even though the sample size is very small, all five cancer patients had potentially resectable disease."
Murtaza envisions a clinical workflow where a patient produces a 40-ml to 50-ml urine sample during a routine visit to their doctor, who then sends the sample to TGen and its collaborators. After extracting cfDNA from 10 ml of the urine sample, Murtaza's team would then perform whole genome library preparation, WGS, and data analysis to determine if the patient is at risk of pre-symptomatic cancer.
Murtaza said the current version of the workflow requires about two to three days to produce a result. However, he acknowledged that his team would still need to build a reference control dataset prior to analysis to decide if the patient's pattern differs significantly enough to determine if the patient will potentially develop early-stage cancer.
While Murtaza said that "it is too early" to decide what the commercial path for the urine-based detection method will look like, his team is interested in working with partners who would commercialize the method. Ultimately, he believes the approach can complement plasma-based liquid biopsy approaches for cancer detection and monitoring.