CD19 chimeric antigen receptor (CAR)-modified T cell therapy has produced impressive results in patients (pts) with CD19+B cell malignancies; however, treatment can be complicated by neurologic adverse events (AEs). The presentation and pathogenesis of neurotoxicity (NT) are incompletely understood.We report a clinical and pathologic study evaluating NT in 133 B-ALL, NHL and CLL pts treated with lymphodepletion chemotherapy and CD19 CAR-T cells (NCT01865617). Neurologic AEs were observed in 53 of 133 pts (40%; 19% grade [gr] 1-2; 16% gr 3; 2% gr 4; 3% gr 5). Most neurologic AEs were reversible. Delirium, headache, and language disturbances were the most frequently observed neurologic AEs, presenting in 66%, 55%, and 34% of pts with NT, respectively. Multivariable analysis showed that higher burden of malignant B cells in marrow, lymphodepletion with cyclophosphamide and fludarabine, and higher infused CAR-T cell dose were associated with increased risk of NT and cytokine release syndrome (CRS).The severity of NT correlated with higher peak concentrations of cytokines that may activate endothelial cells (EC), such as IL-6, IFN-γ, and TNF-α. Pts who developed gr ≥3 NT had more severe CRS with higher CAR-T cell counts in blood and evidence of vascular dysfunction. In severe NT, hypofibrinogenemia with elevated PT, aPTT, and d-dimer were consistent with EC activation. We confirmed in vivo EC activation during NT by demonstrating high concentrations of serum angiopoietin-2 (Ang-2) and von Willebrand Factor (VWF), which are released from Weibel-Palade bodies on EC activation. In vitro, serum from pts with NT induced activation and VWF release from human umbilical vein ECs (HUVECs), which suggests that the finding of a reduced fraction of high molecular weight VWF multimers in vivo during gr ≥4 NT results from consumption on platelets and sequestration on activated EC. Consistent with this explanation, we found reduced activity of the VWF-cleaving protease ADAMTS13 relative to VWF levels and more severe thrombocytopenia in pts with gr ≥4 NT.We considered that cytokine-induced EC activation might alter the integrity of the blood-brain barrier (BBB) during NT. During severe NT, CSF analyses demonstrated increased permeability of the BBB to protein and leukocytes, including CAR-T cells. The gradient between serum and CSF cytokines observed before lymphodepletion was lost during acute NT, consistent with inability of the BBB to shield the CNS from high concentrations of plasma cytokines. Brain vascular pericytes (BVP) play a critical role in vascular and BBB support, and when exposed to high TNF-α or IFN-γ concentrations BVP exhibited stress (increased cleaved caspase-3 and reduced PDGFRβ) and secreted IL-6 and VEGF, further promoting EC activation and BBB permeability.Neuropathologic examination of the brain of a patient with fatal NT demonstrated disrupted endothelium by CD31 immunohistochemistry and EC activation was confirmed by intravascular VWF binding and CD61+ platelet microthrombi. Further evidence of breach of the BBB included red blood cell extravasation from multiple vessels, vascular lesions with karyorrhexis, perivascular CD8+ T cell infiltration, and fibrinoid vessel wall necrosis. CAR-T cells were detected in the CNS.We investigated strategies to reduce the risks of severe NT. Logistic probability curves demonstrated that reduction in CAR-T cell dose to reduce the peak in vivo CAR-T cell blood count was associated with reduced risk of NT, but that the narrow therapeutic index of this approach would lead to loss of anti-tumor efficacy. To maintain CAR-T cell peak counts in blood, we developed a strategy to identify pts early after CAR-T cell infusion who might be at risk of subsequent severe NT and could be candidates for early intervention. Classification tree modeling demonstrated that in the first 36 hours after CAR-T cell infusion pts with fever ≥38.9°C, serum IL-6 ≥16 pg/mL, and MCP-1 ≥1343.5 pg/mL were at high risk of subsequent gr ≥4 NT (sensitivity 100%; specificity 94%). We also investigated whether pts with pre-existing endothelial activation were at higher risk of NT. Before lymphodepletion, pts who developed gr ≥4 NT had higher Ang-2:Ang-1 ratios than those with gr ≤3 NT, indicating that endothelial activation before lymphodepletion or CAR-T cell infusion may be a risk factor for NT that identifies pts who would benefit from a modified treatment regimen. DisclosuresTurtle:Adaptive Biotechnologies: Other: Advisory board; Bluebird Bio: Other: Advisory board; Gilead Sciences: Other: Advisory board; Precision Biosciences: Other: Advisory board; Celgene: Other: Advisory board; Juno Therapeutics: Other: Advisory board, Patents & Royalties, Research Funding. Liles:Juno Therapeutics: Consultancy, Other: Advisory board. Li:Juno Therapeutics: Employment, Equity Ownership. Yeung:Gilead: Research Funding. Riddell:Juno Therapeutics: Equity Ownership, Patents & Royalties, Research Funding. Maloney:Celgene: Other: Advisory board; Kite Pharmaceuticals: Other: Advisory board; Juno Theraapeutics: Other: Advisory board, Patents & Royalties, Research Funding; Roche/Genetech: Other: Advisory board.