Gregor Mendel worked out the rules of inheritance in a monastery garden in the 1860s, crossing pea plants and counting wrinkled versus smooth seeds. Two laws came out of that work. Genes segregate predictably during reproduction. Different traits assort independently. Every biology student learns this before they learn anything else about genetics, because the entire edifice of modern heredity sits on top of it.
Cancer cells don't play by those rules. A growing body of research from Stanford, University College London, and several other labs shows that roughly one in five cancers contains small circles of DNA that float free of the chromosomes, multiply themselves outside the normal rulebook, and get passed to daughter cells through something closer to a raffle than a formal inheritance. These circles go by the name "extrachromosomal DNA," or ecDNA, and they're turning out to be one of the reasons chemotherapy stops working so quickly in so many patients.
Cancer has always been a master of breaking biological rules. A normal human cell will divide somewhere around 50 times before it hits a ceiling called the Hayflick limit, a built-in counter that prevents runaway growth by shortening the protective caps on chromosomes called telomeres until the cell gives up and stops dividing. Cancer cells get around this by reactivating an enzyme called "telomerase" or by using alternative lengthening mechanisms, and the result is a cell line that can divide indefinitely. That was the first rule to fall. Mendel's laws are the next one.
The scale of the problem is larger than researchers initially suspected. When ecDNA was first described in the 1960s, it was filed away as a curiosity. As recently as ten years ago, most textbooks put ecDNA in maybe 2% of tumors. Then the sequencing got better. A 2020 analysis of more than 3,200 cancer patients found ecDNA in a sizeable fraction of aggressive tumors. A larger 2024 study of nearly 15,000 patients across 39 tumor types pegged the rate at 17.1%, and the circles were more common after chemotherapy than before it. In some cancer types the number runs considerably higher. In early-stage esophageal cancer, ecDNA shows up in about 24% of tumors. In late stages, that figure climbs to 43%.
Biology already had precedent for this trick. Bacteria use plasmids, small circular pieces of DNA that ride around outside the main chromosome, to share antibiotic resistance genes between cells. Fungi do something similar. The trick is effective because it decouples a particular useful gene from the slow, orderly process of whole-genome inheritance. If the gene is on a circle, copies can multiply fast, spread fast, and appear in populations that need them. The fact that cancer cells use the same basic strategy to amplify oncogenes and resistance genes suggests this is convergent evolution at the cellular level. When you need fast, unpredictable inheritance, circles are the solution life keeps arriving at.
The Raffle Instead of the Split
The reason this matters comes down to how cells divide. When a healthy cell splits into two, the chromosomes line up on the spindle, each one gets yanked by a structure called a centromere, and the result is two daughter cells with a matched set of genes. It's tidy. It's predictable. Mendel's laws describe exactly this process, and they work because the machinery of mitosis is designed to distribute genetic material fairly.
ecDNA has no centromeres. These are bare rings of genetic code, often carrying a handful of cancer-driving genes called oncogenes, and they have no built-in handle for the cell's sorting machinery to grab. So they do something else. They stick to the outside of chromosomes during division and ride along like passengers who jumped into a taxi at the last minute. Whichever daughter cell the host chromosome ends up in, the ecDNA goes too. But because the hitchhiking is random, one daughter might inherit dozens of these circles while the other gets almost none.
Paul Mischel, a pathologist at Stanford who has spent much of the past decade working on ecDNA, described it to a Stanford reporter as two buses pulling up to the same stop. Passengers grab seats wherever they can. Nothing is assigned. If you don't get on, you don't go anywhere, and the unbound circles drift off into the cytoplasm and vanish.
That unfairness is the whole point. A tumor made of cells with wildly different ecDNA counts is a tumor where every chemotherapy dose hits a population with enormous genetic variety. Kill the cells with low oncogene copy numbers, and the survivors will have high copy numbers. Dose the survivors with a targeted drug, and the next generation shuffles the deck again. You're not really fighting one cancer. You're fighting something that evolves on its feet.
This is why resistance shows up so fast in ecDNA-driven tumors. In glioblastoma, where Mischel first stumbled onto the circles while studying the EGFR gene, patients who responded well to targeted therapy sometimes relapsed within months. When researchers sequenced the relapsed tumors, the EGFR amplifications that had been sitting on ecDNA before treatment were often gone, replaced by a different genetic configuration entirely. The cancer hadn't mutated. It had just reshuffled.
The math of oncogene expression on ecDNA compounds the problem. A normal cell carrying two copies of a gene will produce a certain amount of protein. A cancer cell with that same gene sitting on twenty copies of ecDNA will produce something closer to fifty or a hundred times more, because the circular architecture of ecDNA makes the chromatin more accessible to the cell's transcription machinery. The DNA is literally easier to read. And recent work has shown that ecDNAs cluster together in structures called hubs inside the nucleus, where the enhancers on one circle can activate the genes on another, producing a kind of collective amplification that no individual chromosome could manage on its own. A 2019 paper in Nature documented hubs containing 10 to 100 ecDNA circles in close spatial proximity, all driving transcription of the same oncogene.
A Weakness in the Hitchhiking
The Stanford team, working with collaborators in the UK and elsewhere under a $25 million Cancer Grand Challenges grant, spent several years trying to figure out exactly how ecDNA sticks to chromosomes. A paper published in Nature in late 2025 laid out the answer. The circles contain short DNA sequences called retention elements. These elements bind to proteins sitting on the chromosomes at spots known as mitotic bookmarks, which are normally used by the cell to remember its identity across divisions, so a liver cell stays a liver cell.
Cancer cells have hijacked the bookmark system outright. The retention elements on ecDNA mimic the way promoters and enhancers interact in healthy genes, which means ecDNA is using the cell's own memory-keeping machinery as a ride-along service.
Here's what makes this interesting from a therapeutic standpoint. That binding is a single point of failure. When researchers blocked the interaction between retention elements and mitotic bookmarks in lab-grown cancer cells, the ecDNA failed to partition during division. The circles ended up in the cytoplasm, got degraded, and the cancer cells died because they had lost the oncogene amplifications that were keeping them alive. This is the kind of result that makes drug developers pay attention. A target that is specific to cancer cells, essential for their survival, and absent in healthy tissue is close to the ideal profile.
Several companies are now working on drugs aimed at various parts of this machinery. Boundless Bio, a San Diego company co-founded by Mischel and Stanford geneticist Howard Chang, has two ecDNA-directed therapies in clinical trials. The first, BBI-355, is a CHK1 inhibitor that exploits a different ecDNA weakness: cells with lots of ecDNA have elevated replication stress, and CHK1 is the protein they rely on to handle that stress. Knock out CHK1 and the replication stress becomes lethal. The POTENTIATE trial has been enrolling patients with EGFR or FGFR2 amplifications, and early pharmacodynamic data suggest the drug is hitting its target in both skin and tumor tissue. The main side effects so far are hematological, which is consistent with CHK1 biology and is manageable.
The second drug, BBI-825, targets an enzyme called ribonucleotide reductase that makes the building blocks cancer cells need to keep copying their ecDNA. A third program, BBI-940, aims at a kinesin involved in ecDNA segregation during mitosis and was scheduled for first-in-human trials in breast cancer patients in the first half of 2026. None of these drugs has a mature efficacy readout yet. The proof of concept data is expected within the next couple of years.
The Fast Track to Resistance
Worth sitting with for a moment is what the inheritance patterns of ecDNA actually mean for a person in treatment. If you have a tumor driven by a gene sitting on a normal chromosome, and your oncologist prescribes a drug that targets that gene, the tumor can develop resistance through point mutations, through alternative signaling pathways, through various escape routes that take real time to evolve. Weeks. Months.
If your tumor is driven by a gene sitting on ecDNA, the resistance mechanism is built in. Every cell division is a redistribution of oncogene copies. Rare cells with favorable combinations exist in the population before treatment begins, because random assortment has been generating them continuously. When the drug arrives, those cells survive by definition. The tumor's genetic diversity isn't something it has to manufacture under pressure. It's the starting condition.
This is the kind of problem that explains a clinical observation oncologists have made for years but had trouble mechanistically justifying. Certain cancers, particularly aggressive glioblastomas, some lung cancers, and subsets of sarcomas, seem to respond well to initial treatment and then come back fast and different. You could blame genetic instability in a general way, and people did for decades. But ecDNA gives the phenomenon a specific shape.
The research on Barrett's esophagus makes this particularly vivid. In a study of nearly 300 patients, 33% of people with the precancerous condition who eventually developed esophageal cancer had ecDNA in their precancerous cells. Of the 40 people in the control group whose Barrett's never progressed to cancer, exactly one had ecDNA, and that person died of an unrelated cause. If those numbers hold up in larger studies, ecDNA detection in precancers could become a serious early warning signal, flagging patients who need intervention before a tumor has formed.
What This Doesn't Solve
Being honest about the limits is worth doing. ecDNA is not the only mechanism of cancer drug resistance, and it's not even the dominant one in most tumor types. Point mutations in drug-binding pockets, upregulation of drug-efflux pumps, and epigenetic changes that silence drug-target expression, all of these continue to drive resistance in cancers that have no ecDNA at all. Even in ecDNA-positive tumors, the circles coexist with chromosomal amplifications, deletions, and single-nucleotide changes that each contribute their own share of trouble.
The 83% of cancers that don't contain ecDNA aren't going to benefit directly from CHK1 inhibitors or kinesin degraders aimed at circle biology. And for the 17% that do contain ecDNA, the clinical trials haven't yet shown whether targeting the circles translates into longer survival in a meaningful way. Phase 1/2 data establishes safety and tells you the drug hits its target. It doesn't tell you the patients live longer. That's a Phase 3 question, and it will take years to answer.
There's also the matter of why ecDNA forms in the first place. The circles originate from errors in DNA replication or repair or from spontaneous rearrangements that release a chunk of chromosomal DNA into the nucleus. Blocking ecDNA function in a cancer cell does nothing to stop new ecDNA from being generated in other cells, and the underlying genomic instability that produces the circles in the first place is a broader problem without an easy handle.
Still, the fundamental shift here matters. For most of the modern era of cancer research, the chromosome was the unit of analysis. Tumors were maps of chromosomal gains, losses, and rearrangements. Drug targets were proteins encoded by genes sitting at specific chromosomal addresses. The whole conceptual framework assumed that the inheritance rules worked out in the 1860s applied.
They don't always. A sizeable fraction of the worst cancers, the ones that kill people fastest and resist treatment hardest, are running on a different system. That system has rules of its own, and researchers are now close enough to those rules to start writing drugs that exploit them.
What happens if a glioblastoma patient ten years from now walks into a clinic and gets tested for ecDNA as part of their initial workup, the way they now get tested for EGFR mutations? What happens if the answer comes back positive and there's a specific combination therapy designed to strip the circles out of the tumor cells before chemotherapy even starts? That's not a promise. It's a possibility, and it's less speculative than it was five years ago.
Mendel's pea plants were never going to explain everything. It just took biology a while to find the exceptions that mattered most.