Damage to DNA in cancer cells can lead to pieces breaking off chromosomes and floating away, like icebergs cracking off from a glacier. Just as icebergs are a threat to ships and their crew, these scattered bits of DNA loom large to physicians and cancer patients by having titanic effects on tumor progression and treatment resistance.
Scientists at Sanford Burnham Prebys Medical Discovery Institute and their colleagues published findings May 28, 2026, in Genome Medicine demonstrating significant similarities between samples of tumors featuring these stranded DNA fragments and research models developed using these tumor samples. This close match provides scientists with greater confidence in using these models to learn how to better diagnose and treat cancers with these splintered spots of DNA.
In 1965, researchers first documented what are now known as circular extra-chromosomal DNA elements (ecDNA) due to their round shape. Then, in 1978, scientists found in mice that ecDNA increased cancer cells’ resistance to a chemotherapy drug.
“More recent studies have found that ecDNA occurs quite often, especially in aggressive tumor types,” said Lukas Chavez, PhD , an associate professor in the Cancer Genome and Epigenetics Program at Sanford Burnham Prebys. “And the presence of these castoff DNA loops is linked with worse clinical outcomes.”
Chavez and his team wanted to study how ecDNA affects pediatric brain cancer using a common research model made by grafting tumor cells from a human patient onto a mouse, but the field lacked evidence regarding how closely the resulting mouse model would match the human tumor.
“Research models made this way take a long time and a lot of care to generate, so it was important for us to determine their validity for studying the behavior of ecDNA,” said Rishaan Kenkre, a research associate in the Chavez lab and lead author of the study.
These patient-derived xenograft (PDX) models are generally noted for their ability to mimic human disease to enable both fundamental research and the testing of new potential treatments. To confirm this was the case for cancer cells with ecDNA, the investigators analyzed nearly 300 pediatric tumor samples spanning 31 types of cancer and compared them to PDX models made from each sample.
The scientists found ecDNAs in just under a third of the samples. The ecDNAs within these models held extra copies of genes that may cause cancer, known as oncogenes. And which oncogenes had additional copies reflected patterns found in large studies of thousands of pediatric tumors.
Next, the scientists narrowed in on a group for which they could access genome sequencing data for both the human tumor sample and resulting PDX models.
“For more than 80% of the PDX models, the presence of ecDNA was consistent with their primary tumors,” said Chavez, the senior and corresponding author of the manuscript. “The ecDNA sequences were largely the same across these pairs as well.”
After also confirming similar numbers of extra oncogene copies in primary tumors and their respective PDX models, the research team turned to a method enabling them to see the effects of ecDNA on individual cells within tumors.
In one of two pairs of brain tumors and corresponding PDX models studied using single-cell sequencing, almost every cell in the tumor had ecDNAs and so did the PDX model. In the other pair, fewer than one in ten cells had ecDNAs, yet the PDX model featured these ecDNAs in virtually every cell.
“We found that the ecDNA-positive cells from the human tumor, despite comprising only a small minority of the tumor population, exclusively grew out into the PDX model,” said Kenkre. “At least for this specific pair, this led us to infer that ecDNAs can provide a selective advantage as the PDX model develops.” These findings further support the idea that ecDNA-positive tumor cells may play a key role in driving tumor growth and recurrence in human patients.
The authors conclude that the similarities they observed between tumors and PDX models suggest that the models are valid tools for continuing to learn about how ecDNAs boost cancer cells, and how to stop it.
The research team plans to use PDX models to study how ecDNAs in cancer cells evolve over time as the cells try to adapt to common treatments such as chemotherapy or radiation.
“Our goal is to gain new insights into ecDNA-associated treatment resistance and reveal new potential therapeutic targets for pediatric cancers,” said Chavez.
Additional authors include:
The study was supported by the National Institutes of Health, National Cancer Institute, National Institute of Neurological Disorders and Stroke Institute, National Science Foundation, Clayes Foundation, St. Baldrick and Dragon Master Foundation.
The study’s DOI is 10.1186/s13073-026-01676-0 .
Genome Medicine
Experimental study
Animals
Preservation and clonal behavior of extrachromosomal DNA in patient-derived xenograft models of childhood cancers
28-May-2026
The authors declare no competing interests.