Glioblastoma kills most patients within 15 months of diagnosis. It is the most aggressive brain cancer in young adults, and for decades, oncologists have treated it as a disease that seems to appear out of nowhere: one scan is clean, the next reveals a fast-growing mass already woven into critical brain tissue. New research from South Korea suggests that picture is wrong. The tumor does not materialize suddenly. It has been building, silently, for years.
A series of peer-reviewed studies led by teams at the Korea Advanced Institute of Science and Technology (KAIST) and Yonsei University has traced glioblastoma’s origins back to a hidden preclinical stage in which cells that look completely normal under a microscope already carry the genetic mutations that will eventually produce a lethal tumor. The work, published across several high-impact journals between 2018 and early 2026, reframes the disease as a slow-burning biological process with a long, quiet fuse, and it raises a question that neuro-oncologists have rarely been able to ask: could glioblastoma be intercepted before it becomes untreatable?
A cancer that starts in hiding
The key location is the subventricular zone, or SVZ, a narrow ribbon of tissue that lines the brain’s fluid-filled ventricles. The SVZ is one of the few places in the adult brain where neural stem cells continue to divide throughout life. That constant activity makes it fertile ground for mutations to take root.
In a landmark 2018 study published in Nature, researchers showed that low-level driver mutations could be detected in SVZ cells located far from the main tumor in glioblastoma patients. The mutant cells appeared capable of migrating through the brain and seeding malignant growth at a distant site. The implication was striking: by the time a radiologist spots a glioblastoma on an MRI, the disease has already traveled from its birthplace.
Building on that foundation, a study in Cancer Discovery examined tissue samples from brain regions that appeared completely tumor-free. Inside that normal-looking SVZ tissue, the researchers found cells already harboring the driver mutations associated with glioblastoma. More importantly, the team mapped stepwise transcriptional programs in those cells, showing that the chaotic mix of cell types found inside a full-blown tumor is not random. It follows a traceable biological sequence that begins well before any visible mass forms. The cellular diversity that makes glioblastoma so resistant to treatment is, in effect, baked in from the start.
A retrograde imaging analysis of IDH-wildtype glioblastoma, the subtype that accounts for roughly 90 percent of diagnoses, put a rough clock on the process. By working backward through serial brain scans of patients who were later diagnosed, researchers estimated that the biological onset of the tumor predates diagnosis by several years. (Note: this link points to the PMC record listed by the original research team; readers should verify that the PMC record corresponds to the described retrograde imaging study, as independent confirmation of the specific PMC ID was not available at the time of publication.) The gap between the moment a mutation first takes hold and the moment a patient walks into a clinic with headaches or seizures is not weeks. It is years.
A related but distinct line of research, published in Science, examined IDH-mutant gliomas, a slower-growing class of brain tumor. Using spatial transcriptomics, a technique that maps gene activity to precise locations within a tissue slice, that team pinpointed cortical glial progenitor cells as the site where the first cancer-driving mutation takes hold. (A direct DOI or stable URL for this Science paper was not available for inclusion at the time of publication; readers can search the Science archives for the KAIST-Yonsei spatial transcriptomics study on IDH-mutant glioma origins.) IDH-mutant gliomas differ from IDH-wildtype glioblastoma in prognosis and biology, but the underlying principle is the same: a single mutation lodges quietly in a normal-looking progenitor cell, and everything that follows is a slow cascade toward cancer.
Piecing together the timeline
Taken together, these studies outline a coherent sequence. First, rare neural stem or progenitor cells in the SVZ or cortex quietly acquire a driver mutation. Second, those cells or their offspring migrate outward and accumulate additional genetic and epigenetic changes over months or years. Third, a critical threshold is crossed and a clinically detectable tumor emerges, already armed with the cellular diversity that blunts surgery, radiation, and chemotherapy.
That timeline matters because it suggests a window of opportunity. If the disease spends years in a covert phase, there is, at least in theory, a period during which intervention could change the outcome. No one has yet built the tools to exploit that window, but knowing it exists is a necessary first step.
What remains uncertain
The emerging model is compelling, but several gaps remain.
The retrograde imaging study that estimated a years-long preclinical phase relied on serial MRI data reviewed after the fact. Raw longitudinal datasets and exact interval calculations have not been made publicly available beyond summary estimates, which limits the ability of independent groups to verify the precise timeline or test whether it holds across different patient populations and age groups.
The evidence for SVZ mutations in humans comes from a limited number of tissue samples. Much of the mechanistic work described in Cancer Discovery draws on a spontaneous somatic mouse model. Mouse brains share key features with human brains, but they are not identical, and the frequency at which these precancerous mutations appear in human SVZ tissue across a broad population has not been established. It is entirely possible that some fraction of healthy people carry low-level driver mutations that never progress to cancer, a phenomenon well documented in blood cancers under the label “clonal hematopoiesis of indeterminate potential.”
The spatial transcriptomics data from the Science paper on IDH-mutant gliomas has not been fully released in publicly accessible repositories. Until count-level data is available for independent reanalysis, the cortical origin claim rests primarily on the published figures and the peer-review process at Science.
Perhaps the most pressing unknown is individual risk. Even if precancerous SVZ cells could be detected reliably, researchers do not yet know which molecular signatures predict progression to glioblastoma and which mark a stable, non-threatening clone. Without that distinction, any screening strategy would risk overdiagnosis, labeling people as pre-cancer carriers when their mutations might never cause harm.
No clinical tool yet, but a shift in thinking
No blood test, cerebrospinal fluid assay, or specialized MRI protocol has been validated to detect these early SVZ or cortical lesions in people who feel well. The mutant cells identified so far are rare and often embedded in tissue that would never be removed during routine care. Picking them up noninvasively would require biomarkers that are both exquisitely sensitive and highly specific, a standard that current liquid-biopsy and imaging technologies have not met for brain tumors.
Researchers in the field have noted that circulating tumor DNA, which has shown promise in cancers of the lung, colon, and breast, faces a particular challenge in the brain: the blood-brain barrier limits how much tumor-derived genetic material leaks into the bloodstream, making detection far harder. Cerebrospinal fluid sampling is more promising but invasive, requiring a lumbar puncture, and no large-scale study has yet tested whether it can reliably flag precancerous SVZ clones.
For now, the practical impact of this research lies more in how scientists and clinicians think about glioblastoma than in how they treat it today. The recognition of a long, silent preclinical phase opens the door to new clinical trial designs focused on intercepting the disease earlier in its evolution, perhaps in patients with incidental brain findings or in those undergoing surgery for unrelated conditions. It also underscores the importance of open data: without broader access to imaging archives and single-cell datasets, independent teams cannot fully test or refine the emerging model.
A years-long fuse that science is only now learning to see
Glioblastoma has long been treated as a bolt from the blue, a cancer that announces itself only when it is already too late. The South Korean research challenges that narrative at its core. The disease is not an overnight catastrophe. It is a years-long biological process that reveals itself only at the very end, after mutated cells have migrated, diversified, and crossed a threshold that current medicine cannot yet detect in time.
Understanding that hidden span, the period when mutated cells still look normal and the brain still functions well, may eventually be the key to shifting glioblastoma from nearly always fatal to something that can be caught early and controlled. That day is not here yet. But for the first time, researchers can describe the fuse. The next challenge is learning to spot it before it burns down.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
