Insights into cytokine release syndrome and neurotoxicity after CD19-specific CAR-T cell therapy

Abstract

T-cells engineered to express CD19-specific chimeric antigen receptors (CD19 CAR-T cells) can achieve high response rates in patients with refractory/relapsed (R/R) CD19+ hematologic malignancies. Nonetheless, the efficacy of CD19-specific CAR-T cell therapy can be offset by significant toxicities, such as cytokine release syndrome (CRS) and neurotoxicity. In this report of our presentation at the 2018 Second French International Symposium on CAR-T cells (CAR-T day), we describe the clinical presentations of CRS and neurotoxicity in a cohort of 133 adults treated with CD19 CAR-T cells at Fred Hutchinson Cancer Research Center, and provide insights into the mechanisms contributing to these toxicities.

Introduction

T-cells engineered to express CD19-specific chimeric antigen receptors (CD19 CAR-T cells) can achieve high response rates in patients with refractory/relapsed (R/R) CD19+ hematologic malignancies. Nonetheless, the efficacy of CAR T-cell therapy can be offset by toxicities, such as cytokine release syndrome (CRS) and neurotoxicity, which could hamper its widespread clinical application.

Efficacy of CD19 CAR-T cell therapy

CAR-T cell therapy has shown remarkable efficacy in patients with R/R CD19+ hematologic malignancies. In R/R ALL, we [1] and others [2–7] have reported minimal residual disease (MRD)-negative complete remission (CR) rates ranging from 60% [5] to 93% [1]. In R/R non-Hodgkin lymphoma (NHL), CAR T-cell therapy achieved best overall response rates (ORR) of 53% to 82% [8–12]; in R/R CLL patients, we [13] and others [14] have reported ORR of 74% and 57%, respectively.

CRS and neurotoxicity: clinical presentation and incidence

Upon recognition of CD19-expressing cells (normal B-cells or CD19+ tumor cells) CD19 CAR-T cells proliferate, exert cytotoxic effects against their target cells, and release cytokines that may trigger a systemic inflammatory response. This is thought to initiate CRS, which is characterized clinically by a variety of systemic symptoms and signs, such as fever, hypotension, capillary leak, coagulopathy and occasionally multiorgan failure. The presentation of CRS may differ between distinct CAR-T cell products, but generally occurs within a few days after CAR-T cell infusion. In most clinical trials of CAR-T cell therapy for ALL, the incidences of CRS and severe CRS were above 70% and 15%, respectively [1, 2, 4–6, 15, 16]. However, in NHL there may be more variability in the incidence of CRS, which was reported at 24% in the TRANSCEND trial [10] and 94% in the ZUMA-1 trial [12]. Comparable anti-tumor efficacy was noted in these trials. Additional studies will be required to confirm these findings and to investigate reasons for the apparent differences in the incidence of CRS.

The incidence of neurotoxicity, defined as any neurological symptoms occurring after CAR-T cells in the absence of another cause, also varies across studies, ranging from 13% [4] to 63% [7] in ALL, and from 7% [10] to 31% [12] in NHL. Neurotoxicity commonly presents with delirium, headache, decreased level of consciousness, or speech impairment. Focal neurologic deficits, seizures, and acute cerebral edema have been reported but are infrequent. Neurotoxicity usually develops after the onset of CRS and can present after its resolution. In almost all cases, neurotoxicity and CRS are reversible.

The FHCRC experience

Recently, we reported a clinical and biological description of CRS [17] and neurotoxicity [18] in 133 patients who underwent CD19 CAR T-cell therapy at our institution to treat ALL, NHL or CLL. The patients received cyclophosphamide +/− fludarabine-based lymphodepletion followed by the infusion of CD19 CAR-T cells formulated in a defined 1:1 ratio of CD4+:CD8+ CAR-T cells. CRS with fever preceded neurotoxicity in a majority of cases. The severity of neurotoxicity was associated with the severity of CRS and accordingly, an early onset of CRS predicted subsequent development of more severe neurotoxicity. CRS of any grade after CD19 CAR-T cell therapy occurred in 70% of patients. The median time to first fever was 2.2 days. We did not observe significant differences in the incidence, severity and clinical presentation of CRS between ALL, NHL and CLL patients, Underscoring the favorable outcome in most patients, only 20% of patients with CRS and/or neurotoxicity received tocilizumab and/or dexamethasone, and CRS was quickly reversible in a majority of cases with a median duration of three days. Ten patients (7.5%) developed grade ≥ 4 CRS and five of these patients died from complications associated with CRS and/or neurotoxicity. All patients with grade ≥ 4 CRS developed non-neurologic organ toxicities. Importantly, a majority of the grade ≥4 CRS and/or neurotoxicity observed in our study occurred during dose finding. Since completion of dose finding, the incidence of these toxicities has fallen.

Higher bone marrow CD19+ tumor burden, severe thrombocytopenia, manufacturing of CAR-T cells from bulk CD8+ T cells rather than central memory CD8+ T cells, Cy/Flu lymphodepletion, and higher CAR-T cell dose were independently associated with all grades of CRS. Severe CRS was associated with higher and earlier CAR-T cell expansion in the blood.

The association between CAR T-cell expansion and toxicity could narrow the therapeutic index of CD19 CAR-T cell therapy. While decreasing the dose of CAR-T cells could prevent toxicity, this could be at the price of a diminished anti-tumor effect. To identify a therapeutic window of CAR-T cell dose leading to maximal efficacy with minimal toxicity, we used logistic regression to model the relationship between peak CAR-T cell counts in the blood and the probability of toxicity, as well as the relationship between peak CAR-T cell counts in the blood and disease response. We observed differences in the shape of these probability curves for disease response depending on the disease subtype. For bone marrow complete response in ALL, the curve was sigmoid, reaching a plateau at approximately 10 CD8+ CAR-T cells/uL, indicating that peak CD8+ CAR-T cell counts > 10 cells/uL might have provided only limited improvement in efficacy. In contrast, the logistic regression analysis showed an increase in CD8+ CAR-T cells beyond 10 cells/uL would likely increase the risks of severe neurotoxicity and CRS. Similar findings were noted for CD4+ CAR-T cells. In NHL patients, the probability curve for complete response did not plateau. Thus, any decrease in peak CAR-T cell counts to decrease toxicity would likely be accompanied by a decrease in efficacy. Other strategies to mitigate toxicity should be investigated in NHL, such as early interventions in patients at high risk for severe CRS. Using classification tree modeling we found that patients with high fever (≥ 38.9°C) and serum MCP-1 concentrations ≥ 1343.5 pg/mL within 36 hours of CAR-T cell infusion were at high risk of developing grade ≥4 CRS. This approach could predict grade 4–5 CRS with high sensitivity (100%) and specificity (84%). We also designed another model predicting grade 4–5 neurotoxicity that incorporated these parameters and serum IL-6 ≥ 16pg/ml. These approaches need to be validated to predict toxicity associated with other CAR-T cell products, in which the kinetics of cytokine levels and clinical presentation may differ.

Pathogenesis of CRS and neurotoxicity

The pathogenesis of CRS and neurotoxicity is incompletely understood. Analysis of patients treated in our study provided insights into the mechanisms contributing to these toxicities [17, 18]. The presence of hypotension, capillary leak, and a consumptive coagulopathy suggested that endothelial dysfunction might be present in severe CRS and/or neurotoxicity. We examined serum angiopoietin-2 and -1 and von Willebrand factor (VWF) concentrations in patients after CAR-T cell infusion and found evidence of endothelial activation during severe CRS and severe neurotoxicity. Serum from patients with severe toxicity contained high concentrations of endothelium-activating cytokines, and activated HUVEC cells in vitro, leading to binding of VWF and platelets. In some patients endothelial activation preceded lymphodepletion and CAR-T cell therapy, indicating that pre-existing endothelial activation might increase the risk of subsequent severe toxicity. We found increased permeability of the blood-brain barrier (BBB) in neurotoxicity and found that the BBB failed to protect the CSF from high concentrations of systemic cytokines. Furthermore, some cytokines could activate and induce stress in pericytes - key supporting components of the BBB. Evidence to support the role of endothelial activation and vascular dysfunction was found in autopsy data from a patient with fatal neurotoxicity, in whom we observed a multifocal thrombotic microangiopathy. It remains to be determined whether this mechanism contributes to the majority of cases of mild and reversible neurotoxicity. Development of robust animal models of neurotoxicity and CRS will be critical to elucidate the mechanisms responsible for severe toxicities after CAR-T cell therapy.

Conclusions

CRS and neurotoxicity after CD19 CAR-T cell therapy are usually of mild-moderate severity and reversible. However, some patients develop severe toxicity associated with robust CAR-T cell expansion and high cytokine concentrations in serum. The finding of endothelial activation in these patients may open new avenues to treat or prevent CAR-T cell associated CRS and/or neurotoxicity.