A treatment-resilient lipid metabolic program drives temozolomide resistance in glioblastoma
Temozolomide (TMZ) resistance remains a major barrier to durable control of glioblastoma (GB). Our previous studies showed that simvastatin can enhance TMZ-induced cell death in non-resistant GB cells by disrupting autophagosome-lysosome fusion and engaging stress-response pathways, whereas established TMZ-resistant cells maintain impaired autophagy flux and remain refractory to TMZ, simvastatin, and their combination. Here, we asked whether this loss of therapeutic responsiveness is accompanied by a treatment-resilient lipid state. Targeted LC-MS quantified 304 lipid species across 25 analytical classes in non-resistant (NR) and TMZ-resistant (R) U251-mKate cells under control, simvastatin (ST), TMZ, and TMZ-ST conditions. Global heatmap, principal-component, volcano, and exploratory PLS-DA analyses demonstrated persistent NR/R lipidomic separation across all four treatment states. Resistant cells recurrently displayed enrichment of lysophospholipids, selected sphingolipids/glycosphingolipids, and cholesteryl esters, with depletion or redistribution of several glycerophospholipid and diacylglycerol pools. A family-resolved analysis across 19 lipid families showed that resistance status was most strongly associated with phosphatidylinositol (PI; PERMANOVA R^2=0.52), phosphatidylglycerol (PG; R^2=0.49), lysophosphatidylcholine (LPC; R^2=0.46), lysophosphatidylethanolamine (LPE; R^2=0.45), ether/plasmalogen phosphatidylcholine (R^2=0.45), and phosphatidylcholine (R^2=0.41). Structure-informed target prediction of discriminant lipids generated pathway hypotheses involving Rap1, PI3K-Akt, phospholipase D, calcium, PPAR, and lipid-metabolic signaling. Transmission electron microscopy showed persistent vesicle-rich, autophagosome-like architecture in resistant cells across treatment conditions, consistent with the previously established late-stage autophagy defect. These data extend the autophagy-cholesterol model of TMZ resistance to a broader membrane-remodeling program and indicate that failure of statin sensitization is associated with coordinated lipid-family remodeling rather than a single lipid species or pathway. The identified lipid families and cholesterol-storage phenotype provide testable vulnerabilities for future functional validation. An exploratory family-level linear SVM analysis provided an orthogonal proof-of-concept: several membrane and storage-lipid families retained complete NR/R separation when an entire treatment was withheld, but these internal results were interpreted as supplementary evidence rather than as a validated classifier.