The Lymphoma Tumor Microenvironment Influences Toxicity after CD19 CAR T Cell Therapy

Introduction: Of patients receiving CD19 CAR T cell therapy for large B cell lymphoma (LBCL), approximately 1 in 10 experience severe cytokine release syndrome (CRS) and 1 in 3 experience severe neurotoxicity. While CAR T cells trigger the onset of these toxicities, CRS and neurotoxicity are thought to occur as a consequence of activated myeloid cells amplifying cytokine and catecholamine release, thereby stimulating inflammation both systemically and at the blood-brain barrier. However, patient and tumor-related factors that account for differences in the amount of toxicity remain poorly understood.

Methods: Serum cytokine levels were measured on an ELLA point of care device prior to lymphodepleting chemotherapy and throughout inpatient treatment with CD19 CAR T cell therapy (axicabtagene ciloleucel) for LBCL. Catecholamine levels were measured as we have previously reported. Tumor biopsies were taken within 1 month prior to infusion of CAR T cells. RNA expression was measured by RNAseq and/or a Nanostring IO360 panel consisting of 770 genes found in the tumor microenvironment (TME) in cancer. Analysis used nSolver to identify cell types, GSEA and differential gene expression between groups. Mouse CAR T cell studies utilized mouse CD19-targeted CAR T cells derived from C57BL/6 splenocytes and cultured in vitro with myeloid cells and target cells to evaluate cytotoxicity and/or cytokine secretion. Elicited mouse macrophages were collected from peritoneal fluid 4 days after IP injection of 3% Brewer's thioglycollate medium. In vivo studies with mouse CD19-targeted CAR T cells were performed in IL2Ra-/- mice given cyclophosphamide as a pre-conditioning chemotherapy followed by adoptive transfer and analyses for CAR T cell and B cell persistence, as well as cytokines.

Results: Of 58 patients undergoing CD19 CAR T cell therapy for LBCL, 8 (14%) had severe (grade 3 or higher) CRS and 16 (28%) had severe (grade 3 or higher) neurotoxicity. At baseline, peripheral blood levels of IL-6, IFN-γ, IL-15 and ferritin were significantly higher in patients who would subsequently experience severe CRS and severe neurotoxicity. Confirming our recent animal model of CRS we determined that peak serum catecholamine levels were higher in patients experiencing severe CRS. To identify if myeloid cells potentiate cytokine release we co-cultured CAR T cells with CD19 target and macrophages obtained from elicited mouse peritoneum. When these macrophages were added, IL-6 release from CAR T cells significantly increased compared to when macrophages were absent. Next, we studied the baseline TME in LBCL CAR T patients. Of 36 patients, 10 (27%) experienced severe neurotoxicity following CAR T cell therapy. By cell type score, the severe neurotoxicity group had a lower expression of genes associated with T cells overall and specifically Tregs. Also significantly lower in the severe neurotoxicity group were T cell genes including multiple subunits of CD3, CD3ζ, FOXP3, ICOS, CD62L and others.

Association of increased T cell infiltration in the TME with low neurotoxicity raised the possibility that suppressive T cell subsets play a role in limiting toxicity post-CAR T cell therapy. To test this hypothesis, we injected CD19-targeted CAR T cells into an immune competent mouse model of Treg depletion (IL2Ra^-/-) with established CD19+ leukemia. Treg deficient mice experienced a massive cytokine release after CAR T infusion and died prematurely due to CAR T toxicity compared to control mice with Tregs intact.

Conclusions: Our observations suggest that the incidence of severe toxicity following CD19 CAR T cell therapy is influenced by baseline characteristics that are present prior to the infusion of CAR T cells. These include systemic inflammation characterized by high cytokine levels and a TME notable for a lack of infiltrating T cells. We posit a model whereby inflammation primes myeloid cells that are further activated upon CAR T cell infusion to release toxic amounts of cytokines and catecholamines. T cell subsets in the TME may modulate CAR T cells at the site of antigen encounter and prevent excessive CAR T activation. Reducing systemic inflammation or encouraging T cell infiltration into tumor prior to CAR T infusion are potential strategies for lowering the toxicity associated with CAR T therapy.