Abstract Diffuse midline gliomas (DMGs) and other high-grade gliomas (HGG) are the most commonly lethal intrinsic brain tumors of childhood. Accordingly, novel treatment approaches are needed that consider the heterogeneity of these tumors. Using an extensive pharmacological screen, we identified several two-drug combinations that were particularly effective against a range of DMG and HGG cell lines, the most active being the combination of the histone deacetylase inhibitor, panobinostat, and a proteasome inhibitor, marizomib, both in clinical trials, which synergistically induced tumor-specific apoptosis. However, at clinically achievable concentrations, a small population of resistant tumor cells survived and dominated, mimicking the clinical scenario, which prompted our development of drug resistance models as mechanistic tools. Resistance mechanisms were examined using RNA sequencing of drug-resistant versus -naïve cells. NMNAT2, a critical mediator of NAD synthesis, was overexpressed in DIPGs in the context of high baseline QPRT expression, suggesting a critical role of NAD synthesis and utilization pathways in resistance. Gene-set enrichment analysis suggested a key role for glycolytic pathways in the resistance phenotype. In support of this hypothesis, we observed increased levels of mitochondrial mass and NAD and ATP in resistant cells compared to drug-naïve cells, as well as upregulation of glycolytic intermediates by LC/MS. Conversely, inhibition of OXPHOS and glycolysis pathways resensitized resistant glioma cells to apoptosis in vitro. To test this hypothesis in vivo, mice were injected with drug-naïve or resistant DIPG-013 glioma cells and were exposed to an OXPHOS and glycolysis inhibitor, lonidamine. Lonidamine alone extended survival of naïve (18.6 + 1.4 d vs. 25 + 1.6 d, p=0.0003) and resistant (18.7 + 1 d vs. 62 + 1 d, p=0.0002) DIPG13p models (Log-Rank). This supports the application of metabolic therapy as a novel strategy for treating resistant MGs that has activity in a clinically relevant DIPG model. Supported by NIHS10OD023402, the DIPG/DMG Collaborative, the CNI, and Connor’s Cure.