Automated Glioblastoma Organoid Platform Reveals Vulnerability to Proteasome and HDAC Inhibitors
Key Insights
A high-throughput drug screening platform using patient-derived glioblastoma (search) tumor organoids tested 166 FDA-approved oncology drugs within eight days of surgery, revealing that only 0.048% of drugs reduced viability to ≤50%.
Proteasome inhibitors (bortezomib, ixazomib, carfilzomib) and HDAC inhibitors (romidepsin, panobinostat) demonstrated the most potent anti-glioblastoma (search) activity, with favorable Cmax/IC50 ratios for carfilzomib and romidepsin.
The brain-penetrant proteasome inhibitor marizomib showed limited ex vivo potency (>10 µM IC50 in most cases), potentially explaining its failure in the recent EORTC-1709 phase III trial.
A research team has established an automated high-throughput drug screening platform based on standardized patient-derived glioblastoma (search) tumor organoids (TOs), performing a drug screen of 166 FDA-approved oncology drugs within eight days after surgery. The study, published in npj Precision Oncology, reveals that glioblastoma exhibits striking vulnerability to proteasome inhibitors and HDAC inhibitors, while demonstrating broad resistance to most other drug classes.
The platform addresses a critical gap in glioblastoma (search) research: most preclinical drug response data have been obtained from GBM cell lines or glioma stem-like cells that lack the tumor microenvironment. Patient-derived tumor organoids preserve the parental genotype and phenotype, maintain cellular heterogeneity, and retain key components of the tumor microenvironment, making them more representative avatars of patient tumors for drug discovery.
A standardized, rapid organoid model
The researchers developed a protocol to generate standardized glioblastoma (search) TOs in large quantities from freshly resected tumor tissue. Single cells re-aggregated to form organoids within one to two days, with TO size remaining comparable among biological replicates and morphologically stable throughout a 10-day observation period. The organoids mainly consisted of viable cells and expressed glial fibrillary acidic protein (GFAP) in the majority of cells, indicating the presence of tumor cells, alongside strong expression of the extracellular matrix protein Tenascin C (TNC), one of the most highly upregulated genes in GBM.
Screening reveals narrow therapeutic vulnerabilities
Drug screening was conducted on TOs from nine patients with newly diagnosed GBM and two with recurrent GBM. The 166 FDA-approved drugs were tested at a single concentration of 2.5 µM in triplicate for 72 hours, with dose-response curves generated for the top five drugs from each case.
The large-scale screen revealed that only a small fraction of drugs reduced TO viability to or below 50%—just 0.048% (n = 8/166) across all patients, with substantial inter-patient differences. The most sensitive drugs were bortezomib (21.8% average viability, range 14.2–49.7), ixazomib (24.1%, 7.4–74.4), carfilzomib (24.2%, 12.9–68.6), romidepsin (24.9%, 6.4–71.0), panobinostat (31.5%, 10.2–57.8), idarubicin (41.8%, 23.4–71.4), omacetaxine (47.8%, 17.1–103.3), and daunorubicin (47.8%, 17.4–74.1).
Hierarchical clustering grouped drugs with similar modes of action, providing evidence for drug class-specific vulnerabilities. The most sensitive drug group consisted of proteasome inhibitors (ixazomib, carfilzomib, bortezomib), followed by epigenetic regulators (panobinostat, romidepsin), and topoisomerase inhibitors (doxorubicin, idarubicin, daunorubicin, epirubicin). No significant differences in drug responses were observed between primary and recurrent GBMs, though this finding was limited by the small sample size of recurrent tumors (n = 9 vs n = 2).
Favorable pharmacokinetic ratios prioritize specific agents
Dose-response curves revealed that the proteasome inhibitor carfilzomib was among the most effective drugs in all 10 evaluable cases. Bortezomib and ixazomib were effective in 80% of patients. The average IC50 values were 91 nM for carfilzomib (51–168, 95% CI), 64 nM for bortezomib (25–153), and 38 nM for ixazomib (20–68). The HDAC inhibitor romidepsin demonstrated IC50 values below 3 nM (0.01–14), while panobinostat averaged 259 nM (77–049).
When comparing IC50 values to peak serum concentrations (Cmax), favorable Cmax/IC50 ratios were achieved for carfilzomib (ratio 64.62x), romidepsin (232.33x), and to a lesser extent bortezomib (4.88x) and ixazomib (3.11x). In contrast, panobinostat (0.32x) and idarubicin (0.05x) appeared unlikely to reach sufficient drug levels.
Target validation confirms therapeutic vulnerabilities
RNAi-mediated knockdown experiments in GBM cell lines (NCH82 and NCH89) confirmed the relevance of the identified targets. PSMB5 (search) depletion by two different siRNAs significantly decreased cell viability in both cell lines. While single knockdown of HDAC1 (search) or HDAC2 (search) had no effect, combined HDAC1/2 double knockdown significantly reduced viability, consistent with romidepsin's specificity for HDAC1/2.
Marizomib's limited potency may explain clinical failure
Given that carfilzomib and bortezomib poorly penetrate the blood-brain barrier, the researchers tested marizomib—a novel, irreversible, brain-penetrant pan-proteasome inhibitor that reaches approximately 30% of its serum drug level in the CNS. However, marizomib proved ineffective in the majority of cases, with combined IC50 values exceeding 10 µM. The Cmax of marizomib is estimated at up to 182 nM, translating to a conservatively calculated Cmax/IC50 ratio of 0.018, and only 0.005 when accounting for 30% BBB penetration.
These findings align with the recent EORTC-1709 phase III trial, in which adding marizomib to standard temozolomide and radiation did not improve overall survival or progression-free survival in newly diagnosed GBM patients. The study authors note that "insufficient ex vivo potency relative to achievable systemic exposure may have contributed to the limited clinical activity of marizomib," though they caution this remains hypothetical given that the TO model does not recapitulate blood-brain barrier physiology.
Toward personalized medicine in glioblastoma (search)
The entire drug screening protocol can be completed within eight days of surgery, providing a feasible point-of-care testing timeframe in a clinical setting—substantially faster than other protocols requiring weeks to months. The researchers acknowledge that Cmax values derived from systemic plasma exposure do not fully capture drug levels in tumor compartments, and that Cmax/IC50 ratios should be interpreted as surrogate parameters. They recommend that candidates with favorable ratios be prioritized for follow-up studies in orthotopic GBM models with measurements of brain and intratumoral exposure.
Future prospective studies should evaluate whether patient-individual organoid drug responses can guide therapeutic decision-making in GBM, ideally combined with matched genomic, transcriptomic, and single-cell profiling of parental tumors and corresponding organoids.