Brain capillary obstruction during neurotoxicity in a mouse model of anti-CD19 chimeric antigen receptor T-cell therapy

Immunotherapy for haematologic malignancies with CD19-directed chimeric antigen receptor T cells has been highly successful at eradicating cancer but is associated with acute neurotoxicity in ∼40% of patients. This neurotoxicity correlates with systemic cytokine release syndrome, endothelial activation and disruption of endothelial integrity, but it remains unclear how these mechanisms interact and how they lead to neurologic dysfunction. We hypothesized that dysfunction of the neurovascular unit is a key step in the development of neurotoxicity. To recapitulate the interaction of the intact immune system with the blood–brain barrier, we first developed an immunocompetent mouse model of chimeric antigen receptor T-cell treatment-associated neurotoxicity. We treated wild-type mice with cyclophosphamide lymphodepletion followed by escalating doses of murine CD19-directed chimeric antigen receptor T cells. Within 3–5 days after chimeric antigen receptor T-cell infusion, these mice developed systemic cytokine release and abnormal behaviour as measured by daily neurologic screening exams and open-field testing. Histologic examination revealed widespread brain haemorrhages, diffuse extravascular immunoglobulin deposition, loss of capillary pericyte coverage and increased prevalence of string capillaries. To measure any associated changes in cerebral microvascular blood flow, we performed in vivo two-photon imaging through thinned-skull cranial windows. Unexpectedly, we found that 11.9% of cortical capillaries were plugged by Day 6 after chimeric antigen receptor T-cell treatment, compared to 1.1% in controls treated with mock transduced T cells. The capillary plugs comprised CD45+ leucocytes, a subset of which were CD3+ T cells. Plugging of this severity is expected to compromise cerebral perfusion. Indeed, we found widely distributed patchy hypoxia by hypoxyprobe immunolabelling. Increased serum levels of soluble ICAM-1 and VCAM-1 support a putative mechanism of increased leucocyte–endothelial adhesion. These data reveal that brain capillary obstruction may cause sufficient microvascular compromise to explain the clinical phenotype of chimeric antigen receptor T-cell neurotoxicity. The translational impact of this finding is strengthened by the fact that our mouse model closely approximates the kinetics and histologic findings of the chimeric antigen receptor T-cell neurotoxicity syndrome seen in human patients. This new link between systemic immune activation and neurovascular unit injury may be amenable to therapeutic intervention.