KAIST chip reveals patient-specific glioblastoma responses in lab tests

Inside a small microfluidic chip, tumor cells from a glioblastoma patient sit beside a model of the blood vessels that surround the tumor. That pairing is the news: a KAIST-led team has created a patient-specific blood–brain tumor barrier chip that showed different treatment responses even when conventional genetic testing predicted the same outcome.

Glioblastoma is difficult to treat partly because the brain's blood–brain barrier blocks harmful substances from entering brain tissue—and can block anticancer drugs too. When a tumor develops, that barrier changes, but not in the same way for every patient. The chip recreates this route into the tumor by co-culturing patient-derived glioblastoma cells with brain vascular endothelial cells and astrocytes, support cells found in nervous tissue. The design can also accommodate perivascular and immune cells.

The researchers tested tumor cells from three patients with temozolomide, or TMZ, and bevacizumab, or BEV, both standard glioblastoma treatments. All three patients had the same result on the MGMT promoter methylation biomarker, so genetic testing suggested they would respond similarly. On the chips, however, the vascular barrier characteristics and drug responses differed from one patient to another.

The chip's results showed a high level of agreement with the patients' actual clinical courses. That suggests the vascular environment may help explain treatment differences that tumor genetics alone does not capture. The findings were published in Small, and the team was led by Professor Song Ih Ahn of KAIST, with collaborators at Sungkyunkwan University, CHA Bundang Medical Center and CHA University.

And practically? If larger studies confirm the result, doctors and researchers could test several drugs on a chip made from a patient's own tumor cells and barrier, helping identify promising options in advance. The same platform could also screen new drug candidates. For now, it is a laboratory research platform, not a validated clinical decision tool: the team says it still needs testing in a larger patient population for predictive performance and reproducibility.