Glioblastoma is one of the most aggressive and difficult-to-treat brain cancers, yet its response to therapy can vary dramatically from one patient to another. A drug that slows tumor growth in one person may have little effect in another, even when both patients appear to have similar genetic profiles and receive identical treatment. Researchers in South Korea have now developed a patient-specific “blood–brain tumor barrier on a chip” designed to reproduce not only a patient’s tumor cells, but also the living vascular environment that determines how drugs reach and affect the cancer. The platform could offer a new way to predict treatment responses before therapy begins and may eventually support more personalized drug selection for people with glioblastoma.
The device was developed by a team led by Professor Song Ih Ahn of the Department of Mechanical Engineering at KAIST, in collaboration with researchers at Sungkyunkwan University, CHA Bundang Medical Center, and CHA University. Their work addresses a central problem in glioblastoma treatment: the tumor is not an isolated mass of malignant cells. It exists within a highly complex brain environment shaped by blood vessels, astrocytes, immune cells, and the altered barrier that forms around the tumor. Because anticancer drugs must cross this barrier before reaching tumor cells, the same concentration of a drug may produce very different biological effects depending on how permeable or restrictive that barrier is in an individual patient.
The healthy blood–brain barrier is formed largely by specialized brain endothelial cells that line blood vessels and tightly regulate the movement of molecules between the bloodstream and neural tissue. These cells are supported by astrocytes and other components of the neurovascular unit. Tight junctions between endothelial cells limit the passage of many substances, including numerous therapeutic compounds. When glioblastoma develops, the surrounding vascular system becomes abnormal. The resulting blood–brain tumor barrier may contain regions that are more permeable than a healthy barrier, while other areas remain highly restrictive. This uneven and patient-specific architecture can influence drug delivery, tumor metabolism, resistance mechanisms, and the survival of malignant cells beyond what can be predicted from tumor genetics alone.
To recreate this environment, the researchers constructed a microfluidic chip containing separate but closely connected channels. Brain microvascular endothelial cells, known as HBMECs, were placed in an upper vascular channel, while astrocytes and patient-derived glioblastoma cells were cultured in a lower tissue channel. The arrangement allows researchers to observe how the two compartments interact while measuring the movement of drugs across the engineered barrier. Microfluidic systems use precisely controlled chambers and fluid pathways that can reproduce aspects of blood flow, concentration gradients, and tissue organization on a small scale. The design can also accommodate perivascular and immune cells, creating the possibility of modeling an even broader range of biological interactions within the tumor microenvironment.
The team used tumor cells obtained from three glioblastoma patients to create individual blood–brain tumor barrier models. Each patient had the same IDH-wildtype classification and the same MGMT promoter methylation status, a biomarker commonly used to estimate sensitivity to temozolomide, one of the standard drugs used against glioblastoma. On the basis of conventional clinical testing, the patients might therefore have been expected to respond in broadly similar ways. The researchers then exposed the patient-specific chips to temozolomide and bevacizumab, an antibody that targets vascular endothelial growth factor and is used to influence tumor-associated blood vessels. The experiments allowed the team to evaluate both the ability of the drugs to cross the engineered barrier and their effects on the tumor cells behind it.
The results revealed differences that were not apparent from the shared genetic and biomarker profiles. The three models showed distinct barrier properties, including differences in permeability and electrical resistance, measurements that reflect how tightly the endothelial layer controls transport. They also displayed differences in gene expression and in their responses to the two anticancer agents. In practical terms, the same drug encountered a different biological gateway in each chip. A treatment could therefore be affected not only by whether the tumor cells were intrinsically sensitive to the drug, but also by how efficiently the patient-specific vascular barrier allowed the drug to reach them and how the surrounding cells altered the tumor’s behavior.
The researchers compared the chip findings with the actual clinical courses of the three patients, including progression-free survival and post-progression survival. According to the team, the on-chip barrier characteristics and drug responses closely matched the outcomes observed in the patients. The comparison suggests that the engineered models captured clinically meaningful features of each person’s disease, despite the small number of cases. The result is important because conventional testing often focuses on the cancer cell itself, while the chip evaluates the complete route a treatment must follow: from the vascular channel, across the tumor-associated barrier, and into the tissue containing malignant cells and supporting cells.
The platform could eventually become a tool for testing several therapies from a single patient’s tumor sample before a treatment decision is made. Researchers could expose parallel chips to different drug combinations, doses, or emerging compounds and compare the resulting tumor responses under conditions that more closely resemble the patient’s own disease. Such testing would not replace clinical trials or medical judgment, and the current findings require validation in a substantially larger patient population. The tumors grown on a chip also cannot reproduce every feature of a living brain, including long-range immune responses, systemic drug metabolism, and the full three-dimensional complexity of a patient’s tumor. Nevertheless, incorporating the blood–brain tumor barrier represents a significant advance over models that test drugs only on isolated cancer cells.
The study’s authors say the same approach could also assist pharmaceutical research by providing a human-relevant system for evaluating new drug candidates and studying why treatments fail. The inclusion of perivascular and immune components may make it possible to investigate how blood vessels and immune cells contribute to resistance, invasion, and recurrence. Minsu Ryoo of KAIST and Gaeun Lee of Sungkyunkwan University served as co-first authors. The findings were published in Small on June 27, 2026, and were selected for the journal’s Front Cover. Professor Ahn described the work as a platform for reproducing patient-derived tumor cells together with the blood–brain tumor barrier, with the longer-term goal of developing personalized treatment strategies and a preclinical system for new drug development. If future studies confirm its reproducibility and predictive accuracy, the chip could help move glioblastoma care toward a model in which treatment is selected not only by what a tumor is, but also by the living barrier that determines whether medicine can reach it.
Subject of Research: Patient-specific blood–brain tumor barrier-on-a-chip models for predicting glioblastoma treatment responses
Article Title: Human Blood-Brain Tumor Barrier on a Chip to Investigate Personalized Treatment for Glioblastoma Patients
News Publication Date: 18-Aug-2026
Web References: KAIST; https://doi.org/10.1002/smll.202506712
References: Small, published 27-Jun-2026; DOI: 10.1002/smll.202506712
Image Credits: KAIST
Keywords: glioblastoma, blood–brain tumor barrier, blood–brain barrier, organ-on-a-chip, microfluidics, personalized medicine, temozolomide, bevacizumab, drug response prediction, brain cancer, tumor microenvironment, precision oncology
