Glioblastoma (GBM) is a fatal tumor whose aggressiveness, heterogeneity, poor blood-brain barrier penetration, and resistance to therapy highlight the need for new targets and clinical treatments. A step toward clinical translation includes the eradication of GBM tumor-initiating cells (TICs), responsible for GBM heterogeneity and relapse. By using patient-derived TICs and xenograft orthotopic models, we demonstrated that the selective lysine-specific histone demethylase 1 inhibitor DDP_38003 (LSD1i) is able to penetrate the brain parenchyma in vivo in preclinical models, is well tolerated, and exerts antitumor activity in molecularly different GBMs. LSD1 genetic targeting further strengthens the role of LSD1 in GBM TIC maintenance. GBM TIC plasticity supports their adaptation and survival under a plethora of environmental stresses, including nutrient deficiency and proteostasis perturbation. By mimicking these stresses in vitro, we found that LSD1 inhibition hampers the induction of the activating transcription factor 4 (ATF4), the master regulator of the integrated stress response (ISR). The resulting aberrant ISR sensitizes GBM TICs to stress-induced cell death, hampering tumor aggressiveness. Functionally, LSD1i interferes with LSD1 scaffolding function and prevents its interaction with CREBBP, a critical ATF4 activator. By disrupting the interaction between CREBBP and LSD1-ATF4 axis, LSD1 inhibition prevents GBM TICs from overcoming stress and sustaining GBM progression. The effectiveness of the LSD1 inhibition in preclinical models shown here places a strong rationale toward its clinical translation for GBM treatment.