Optimizing the IFNγ Axis Improves CAR T-Cell Potency in AML but Not B-ALL

Background: Despite the remarkable success of chimeric antigen receptor (CAR) T-cell therapy for treatment of B-cell acute lymphoblastic leukemia (B-ALL), translational to acute myeloid leukemia (AML) has been disappointing. Interferon-gamma (IFN𝛾) is a keystone pro-inflammatory cytokine produced by CAR T cells, universally measured as a benchmark of CAR T-cell activity. The requirement for IFN𝛾 appears to be context- and tumor-dependent, necessary for optimal activity against solid tumors but not against B-cell malignancies. Thus, in B-ALL and B-cell lymphomas, pharmacologic blockade of IFN𝛾 has emerged as an attractive strategy to mitigate CAR T-cell related toxicities without impacting anti-tumor activity. However, the importance of an intact IFN𝛾 axis has not been studied with regards to anti-AML CAR T-cells. Malignant myeloid cells secrete and respond to cytokines in a distinct manner from malignant B cells and therefore the resulting tumor-CAR T cytokine milieus may be different. We hypothesized that CAR T cells require IFN𝛾 to successfully kill AML, and that the soluble factors secreted after AML-CAR T engagement may limit CAR T-cell activity.

Methods/Results: To eliminate target- or CAR-specific differences, we engineered the Nalm-6 B-ALL cell line to express AML target CD123 (N6-123) at a range of antigen densities to mimic its variable expression on AML. We also generated AML cell lines exogenously expressing CD19. CD123 or CD19 CAR T cells generated from healthy human donors were then co-cultured with AML or B-ALL cell lines with their respective target antigen at a range of effector-to-target (E:T) ratios, with or without anti-IFN𝛾 antibody. In a CAR T cell “challenge” experiment, additional tumor cells were added every 24 hours for 4 days and tumor cell growth was measured in an Incucyte cytotoxicity assay. ELISA confirmed that the anti-IFNg antibody effectively blocked IFN𝛾 but did not affect IL-2 secretion. Corroborating published results, pharmacologic IFN𝛾 blockade did not impact anti-leukemic activity of CD123 CAR T cells against B-ALL N6-123 cells. In contrast, disruption of the IFN𝛾 axis significantly impaired the activity of CD123 CAR T cells in the AML setting. Furthermore, CD19 CAR T-cell activity was not dependent on IFN𝛾 in B-ALL but was compromised when targeting AML expressing exogenous CD19. To determine whether this differential IFN𝛾 dependence could be leveraged to augment CAR T-cell activity, AML cell lines were pre-incubated with IFN𝛾, then exposed to CD123 CAR T cells. Indeed, CD123 CAR T cells killed pre-exposed AML cells more effectively than those treated with media alone.

To more broadly assess the differences in the cytokine milieu after AML-CAR T or B-ALL-CAR T engagement, and the impact of IFN𝛾 on the “soluble factor signature” in each tumor type, we performed 31-plex Luminex to measure concentrations of pro- and anti-inflammatory cytokines, chemokines, and growth factors in a 24h co-culture in the same model system described above. We identified a number of cytokines and chemokines that were “AML predominant,” including IL-1Ra and TGFβ, both of which could interfere with CAR T-cell function based on known biologic functions. A subset of these differentially secreted factors were dependent on an intact IFN𝛾 axis, suggesting a possible mechanism for the differential IFN𝛾 dependence in the experiments described above.

Conclusions: We identified a critical role of IFN𝛾 in driving CAR T-cell cytotoxicity against AML, which is in stark contrast to B-ALL and an important discrepancy to consider when devising cytokine-modulating therapies, which could include pre-CAR T cell interventions like chemotherapy conditioning regimens or post-CAR T-cell interventions like targeted immunomodulatory agents to mitigate toxicity. All of these factors must be considered in a tumor-specific context. Furthermore, our data suggest that the IFN𝛾 axis could be exploited to optimize AML CAR T-cell therapeutic windows, particularly for CD123, which is upregulated in the setting of pro-inflammatory cytokines. Finally, we highlight distinct soluble factor signatures in the AML-CAR T vs the B-ALL-CAR T settings, which provides rationale for pharmacologic and/or CAR T-cell engineering strategies to modulate cytokines in order to create a more hospitable hematologic microenvironment for AML CAR T cells.