Preprint: BMP Signaling Shifts Drug-Induced Death in Patient-Derived Glioma Spheroids
An imaging assay separates growth arrest from death in glioma spheroids. BMP signaling protected against selected alkylating agents and kinase inhibitors while increasing sensitivity to lipid-dependent necrosis, including tegavivint-induced death.
Measuring Death Directly in Three-Dimensional Glioma Cultures
A bioRxiv preprint from Stanford University and the University of California, Los Angeles examines why the same glioma cell state can resist some drugs yet become vulnerable to others. The researchers developed HADES, an imaging assay that follows growth and cell death in patient-derived glioblastoma and diffuse intrinsic pontine glioma spheroids. They used it to test how extracellular protein ligands alter responses to a panel of compounds.
HADES combines a nuclear red fluorescent live-cell marker with the dead-cell dye SYTOX Green. Changes in both signals over time generate a spheroid-death score. This provides a different measurement from ATP-based viability assays, which can decrease when cells stop growing as well as when they die.
In glioblastoma spheroids, HADES distinguished lethal effects that the CellTiter-Glo 3D assay did not resolve. Trypan-blue staining supported the imaging results. Several kinase inhibitors all lowered the ATP-associated signal, yet osimertinib and buparlisib caused more measured death than JCN037 or paxalisib. The assay comparison therefore established the distinction needed for the subsequent ligand screen.
A Screen Reveals Opposite Responses to Different Drug Classes
The main screen tested 50 candidate ligands against 13 drugs or tool compounds in a patient-derived glioblastoma line. Ligands were given for 24 hours before compound exposure, and imaging followed responses for five days. The experiment generated 4620 dose-response curves from 50820 individual spheroid measurements. Two autofluorescent ligands were excluded from further analysis; 39 of the 50 tested ligands did not significantly alter death in response to any of the tested compounds under the screen criteria.
Bone morphogenetic proteins 2 and 4, or BMP2/4, produced a reproducible pattern across additional patient-derived lines. They protected spheroids from temozolomide, the alkylating agent VAL-083 and selected kinase inhibitors, while increasing sensitivity to the lipid-dependent-necrosis inducer CIL56. BMP3 did not reproduce the effect. BMP2 also failed to alter drug-induced death responses in the tested adherent T98G cell model despite biochemical pathway activation, emphasizing a dependence on model context.
Receptor inhibitors and SMAD4 gene disruption supported a requirement for canonical BMP–SMAD signaling. The protection against alkylating agents was associated with reduced cell-cycle progression and less drug-associated DNA damage. Pretreatment with the CDK4/6 inhibitor palbociclib reproduced that protective response in a responsive spheroid line. By contrast, BMP2 increased apoptotic priming and sensitivity to combined BCL-2 family inhibition. The effect depended on the type of death stimulus.
Partial BMPR1A silencing did not alter the compound-induced death response. Partial SMAD4 perturbation attenuated BMP2-mediated CIL56 sensitization but did not change the lethality of temozolomide, JCN037 or buparlisib. Three SMAD4-disrupted spheroid populations resisted BMP2-mediated changes across the tested drugs. Disrupting p21 or p27 individually also failed to prevent the growth and death effects.
Lipid Metabolism Creates a Separate Vulnerability
The increased sensitivity to lipid-dependent necrosis followed a different route from growth arrest. BMP2/4 reduced expression of FADS2, an enzyme that consumes palmitate through lipid desaturation. FADS2 inhibition or disruption sensitized spheroids to CIL56 without requiring added BMP. Lipid analyses and palmitoylation perturbations supported a model in which greater palmitate availability enables this non-apoptotic death process.
BMP2 pretreatment also increased sensitivity to tegavivint, a clinical drug candidate related to CIL56. The lipid-dependent-necrosis inhibitor TOFA suppressed death induced by either compound. Palbociclib did not reproduce the lipid-death sensitization, helping separate the cell-cycle and lipid-metabolism arms of the response.
The team extended the assay to patient-derived diffuse intrinsic pontine glioma spheroids. BMP7 protected two ACVR1-wild-type lines against selected drugs while sensitizing them to CIL56 and tegavivint. A line carrying the activating ACVR1 G328E variant had basal SMAD-pathway activity and did not show further ligand-dependent response changes. These models support a related signaling switch, while retaining the different ligand and receptor contexts.
The preprint’s central finding is a tradeoff: reduced proliferation can diminish killing by one drug class while exposing another vulnerability. The proposed strategy of priming tumors for lipid-dependent necrosis remains preclinical. Tegavivint’s usefulness in brain cancer is uncertain, including whether adequate brain exposure is achievable. The mechanism linking growth arrest to reduced kinase-inhibitor lethality and the critical palmitoylated death-execution targets also remain unresolved.