Glioblastoma stem cells, the primary cilium, and a completed study whose answer was never published

What the work claims

This is an interventional, single-group, basic-science study, marked completed with an actual enrollment of 100 participants, running from February 2021 to July 2024. It is not a treatment trial; the primary purpose is listed as basic science. Its central claim is mechanistic: that the cilium-related transcriptome of patient-derived glioblastoma stem cells can be modulated to control tumor growth and invasiveness, that drugs inhibiting cilia disassembly are candidate therapeutics, and that a cilium-related gene expression signature in patient tumor tissue has prognostic value.¹

The registered primary outcome states the test precisely: the correlation between modulation of cilia disassembly complex genes and tumor growth and invasion in brain organoids, measured through study completion over an average of two years. The keyword list names the machinery: cilia disassembly complex, Nek2, glioblastoma stem cells. The intervention modules describe the full arc: collect tumor tissue and blood to build glioblastoma stem cell cultures and brain organoids; dissect ciliogenesis players by transcriptomics and by modulating cilium-related genes with cilia-targeted drugs; then validate the ciliogenesis players as a prognostic signature back in patient tumor tissue.¹

How it works

The primary cilium is a microtubule-based signaling organelle that most vertebrate cells grow when they are quiescent. It concentrates growth-factor receptors, hedgehog pathway components and mechanosensors, and it must be disassembled before the cell can re-enter the cell cycle. Disassembly is executed by a defined protein complex, with the kinase Nek2 as one of its best-known members: Nek2 phosphorylates ciliary proteins including KIF24 to trigger resorption at the G0-to-G1 transition. The study's bet is that glioblastoma stem cells, the therapy-resistant subpopulation thought to drive recurrence, live in and out of a quiescent state, and that their ability to exit quiescence depends on an intact cilia disassembly machine.

If that is true, the pharmacology writes itself. Inhibiting disassembly does not kill the cell; it freezes it. A glioblastoma stem cell that cannot resorb its cilium cannot cycle, cannot proliferate a recurrent tumor, and, the registered outcome implies, cannot invade. This is a state-switching strategy rather than a cytotoxic one, which matters for how it would be combined: a cilia-disassembly inhibitor would be a cytostatic partner layered under or before cytotoxic therapy, aimed at the reservoir that survives it.¹

The organoid's role in this design is the most interesting architectural choice. The study's primary outcome is not measured in mice; it is the correlation between cilia-gene modulation and tumor growth and invasion in brain organoids. Patient-derived glioblastoma stem cells are introduced into neural organoid tissue, and invasion into that human neural substrate becomes the readout. The organoid is not modeling the patient here. It is playing the role of the host brain, the tissue being invaded, which makes it an assay of the tumor's behavior against a human target rather than a portrait of the disease.

Where a skeptic should push

The single most load-bearing assumption is that cilia disassembly is a dependency of glioblastoma stem cells specifically, rather than of cycling cells generally. Cilium resorption is not a tumor-specific process; it is how every dividing cell re-enters the cycle. A drug that blocks it systemically would be expected to hit proliferating tissue everywhere, and the ciliopathy field's catalogue of developmental and renal defects from cilium dysfunction is the standing warning label. Tumor selectivity has to come from somewhere else: overexpression of Nek2 or another complex member in glioblastoma stem cells, a shorter window of vulnerability, or delivery across the blood-brain barrier to a tumor that has disrupted it. None of that is demonstrated in the registry, and the registry is all we have.

Second, the endpoint is a correlation, not an intervention readout with a control arm. Correlation between cilia-gene modulation and growth or invasion in organoids cannot by itself establish that the cilium machinery is causal; off-target effects of the modulation tools and the targeted drugs are exactly the sort of thing a single-group design cannot exclude. The registered design also runs the in vitro, in vivo and ex vivo legs in sequence under one protocol, so a failed leg can be absorbed silently into the completed study.

Third, and most simply: the study is marked completed with 100 actual participants, and no results are posted on ClinicalTrials.gov. The registry record, last updated in March 2025, is the only primary document available. Whatever the correlation turned out to be, the public record does not contain it. A skeptical reader should treat every mechanistic claim in this article as a design claim, stated by the investigators as intent, not as a result.

Cilium state as a brain-organoid drug target

The non-obvious implication for organoid models of human organs is architectural. This study uses the brain organoid as the invaded tissue, not as the disease model: the patient-derived tumor cells are the experimental variable, and the organoid supplies a human neural substrate whose destruction is the quantitative readout. That inversion is a genuinely useful pattern for the organoid field. Most brain-organoid work asks whether the organoid recapitulates the patient's tumor; this design asks whether the patient's tumor can destroy human neural tissue when a specific gene program is perturbed. The second question is the one a preclinical screen actually needs answered, and it turns the organoid from a fidelity problem into an instrument.

The opportunity is a new target class with a built-in resistance logic. Glioblastoma recurrence is driven by cells that survive therapy in a quiescent or stem-like state, and a cytostatic switch that holds those cells out of cycle is a mechanistically rational complement to radiation and temozolomide. If the cilia disassembly complex, Nek2 included, is druggable in brain tissue, an organoid-based invasion assay is the right credentialing step for such drugs: it measures the phenotype that matters, infiltration into neural tissue, in a human context, faster and more scalably than mouse orthotopic models. The prognostic-signature arm adds a second use: if a cilium-related transcriptomic signature in bulk tumor tissue identifies patients whose tumors are cilium-dependent, the assay doubles as a companion diagnostic for the target class.

The threat is the generalization trap, and it is severe. This is one center, one tumor type, one organoid invasion assay, and a target whose normal-tissue toxicity profile is a known field of human disease. Cilium biology also varies with cell state: the same organoid that reads out invasion carries a developmental, fetal-like neural context, and the invasion phenotype of glioblastoma stem cells into fetal-like tissue may not match their invasion of an adult brain. But the sharpest threat is the reporting silence. A completed, registered, hundred-participant mechanistic study with no posted results means the field cannot tell whether the cilium correlation held, and the negative result, if it is one, is invisible. Model-dependent drug targets are exactly the claims that most need posted outcomes, because the next lab will build on the published design whether or not the design worked.

The bottom line

Established by the registry record: a completed, registered, single-center basic-science study of 100 glioblastoma patients tested whether modulating the cilia disassembly complex, with Nek2 named as a keyword, controls glioblastoma stem cell growth and invasion in brain organoids, alongside in vitro and in vivo legs and a prognostic-signature validation in patient tumor tissue. The study finished in July 2024 and no results are posted. Hypothesis: that blocking cilium resorption traps glioblastoma stem cells in a non-cycling, non-invasive state and defines a druggable, prognostically selectable dependency. What would confirm it: posted results showing the registered correlation holds across patient-derived models, with on-target validation of the drugs used and a therapeutic window over normal proliferating tissue. What would break it: the completed study's silence itself, evidence that the effect is an off-target artifact of the modulation tools, or a toxicity profile that makes systemic cilia-disassembly inhibition unmanageable. Until the results appear, the cilium is a promising glioblastoma hypothesis with an unpublished test.