Automated Glioblastoma Organoid Screening Prioritizes Proteasome and HDAC Inhibitors
A two-stage screen tested 166 approved oncology drugs in organoids from 11 glioblastomas. Selected proteasome and HDAC inhibitors showed strong ex vivo activity, while marizomib was weak in most tested models and systemic exposure remained an imperfect proxy for brain delivery.
An Eight-Day Workflow from Surgery to Drug-Response Curves
A Heidelberg University-led study developed an automated workflow for screening drugs in short-term glioblastoma tumor organoids. The investigators wanted a format that could produce enough similarly sized models for broad testing while preserving a clinically relevant turnaround. Their NPJ Precision Oncology report uses fresh surgical material to identify drug-class vulnerabilities and then compares the observed potency with available systemic-exposure data.
The models were generated by reaggregating dissociated cells from viable, contrast-enhancing tumor regions. Necrotic or insufficient specimens were excluded. Organoids compacted within one to two days, expressed glial and matrix-associated markers, and showed increasing ATP-associated activity during characterization. These are short-term reaggregated tumor models rather than long-term cultures selected through repeated expansion.
The screen included nine newly diagnosed and two recurrent glioblastomas. On day two, organoids received 166 FDA-approved oncology drugs at 2.5 μM for 72 hours, with triplicate wells and assay controls. The five strongest drugs for each case then advanced to concentration-response testing. The complete two-stage process produced results within eight days of surgery. One of the 11 cases lacked sufficient cells for the second stage.
Drug-Class Responses Require an Exposure Context
Most compounds did not substantially reduce organoid viability at the screening concentration. The strongest average responses clustered among proteasome inhibitors, histone deacetylase inhibitors and selected topoisomerase inhibitors. There was no observed primary-versus-recurrent response difference, but only two recurrent tumors were included. That small comparison cannot exclude disease-stage effects.
In the 10 cases with concentration-response data, carfilzomib appeared among the five most effective drugs in every case. Bortezomib and ixazomib were represented in 80% of cases, while romidepsin and panobinostat appeared in 70% and 60%, respectively. These frequencies refer to the patient-specific shortlists from the screen.
The authors compared ex vivo IC50 values with reported peak serum concentrations, or Cmax. Carfilzomib, bortezomib, ixazomib and romidepsin had favorable ratios under that calculation; panobinostat and idarubicin did not. The comparison provides a way to deprioritize activity requiring implausibly high systemic exposure, but serum concentration does not determine the concentration reaching brain or infiltrative tumor tissue.
Mechanistic follow-up used two established glioblastoma cell lines because patient material was limited. Depleting the proteasome target PSMB5 reduced viability. Depleting HDAC1 or HDAC2 individually did not, whereas combined depletion did. These results support the drug-class hypothesis in cell lines; they are not genetic validation performed directly in the patient organoids.
A Brain-Penetrant Proteasome Inhibitor Provides a Negative Test
The investigators separately tested marizomib in organoids from seven patients. Although prior work described better central-nervous-system penetration than some other proteasome inhibitors, marizomib had limited ex vivo potency in most of these models. The combined IC50 was above 10 μM, with one particularly sensitive case below 3 nM.
This negative result complicates a simple ranking by either brain penetration or drug class. A compound can reach the brain more readily yet require a higher concentration to affect the tumor cells. Conversely, a potent compound in organoids may fail to reach an effective intratumoral concentration. The study discusses both parts of that exposure–response problem.
The authors relate the weak marizomib response to a previously reported negative clinical trial, but the organoid experiment does not establish why the trial failed. It does not reproduce blood–brain barrier physiology, intratumoral pharmacokinetics or the full effects of combination treatment. The proposed connection remains a hypothesis.
The workflow’s practical contribution is a rapid, patient-specific screen followed by a more detailed potency assessment and an explicit exposure review. Its current limits include tissue-selection bias, technical rather than independent repeated screens from each tumor, and surrogate serum-exposure comparisons. Prospective studies linking organoid results, measured tumor exposure and patient outcomes are needed to determine how the workflow might inform treatment decisions.