Chimeric antigen receptor (CAR) T-cell therapies are revolutionizing the treatment of lymphoid malignancies. Despite highly promising preclinical data and a multitude of phase I trials, however, the efficacy of CAR T-cell therapy for acute myeloid leukemia (AML) has so far been discouraging. Several explanations for this have been previously postulated and/or proven, including that autologous T cells are particularly exhausted in heavily pretreated AML patients and make for poor CAR substrates; antigen expression is highly heterogenous in AML; myeloid antigens are expressed on both leukemia and hematopoietic cells, resulting in adverse on-target, off-tumor effects; and the AML microenvironment is too immunosuppressive for effective immunotherapy targeting. Although each of these issues represents a significant limitation, as a firsthand witness of the disappointing results, it has become clear that something else is amiss. In particular, why do most AML patients experience profound cytokine release syndrome (CRS), implying ample CAR activation and killing, yet fail to show an appreciable leukemia response?
Anand S. Bhagwat, MD, PhD, and colleagues have uncovered a clever yet rational explanation for why CAR T-cell therapies have been failing AML patients, regardless of adequate CAR signaling and target antigen exposure (hint: it’s the cytokines!) The authors conducted a phase I clinical trial of autologous CD123-targeting CAR T cells in adults with relapsed or refractory AML. Twelve patients were infused (following fludarabine and cyclophosphamide lymphodepletion [LD]), with 10 (83%) developing CRS (including two grade 5 CRS events), as well as an array of expected adverse events reported. Three patients responded to therapy, two of whom achieved molecular complete response (CR), while one achieved CR with incomplete count recovery and detectable molecular disease. The median overall survival from infusion was 160 days. These results, while important, are only the beginning of the story.
The remainder of the publication delves into why the CD123-targeting CAR T cells were largely unsuccessful at killing AML in most patients. Product characteristics did not hold the answer, as the manufactured products were predominantly central memory T cells, were not overly exhausted, and demonstrated the capacity for proliferation and killing in vitro. Antigen density did not offer the solution either. At baseline, CD123 was moderately to highly expressed on the AML cells of all patients and was not consistently downregulated or lost after CAR T-cell infusion. This is where the investigators got creative, noting that cytokines have been shown to support AML proliferation in other contexts. Could the cytokines released after LD and CAR T-cell infusion inadvertently promote AML growth and survival?
Turns out, this is exactly what was happening, and here is how they proved it. First, they showed that relative to cells incubated with pre-LD serum, AML cells incubated with serum from peak CAR expansion showed substantially enhanced viability. Second, through extended cytokine profiling, the investigators demonstrated that specific “myeloid-supporting” cytokines released after CD123 CAR T-cell infusion (e.g., interleukin 3, granulocyte-macrophage colony-stimulating factor [GM-CSF], and Fms-like tyrosine kinase 3 ligand) significantly promoted AML survival by reducing apoptosis. Importantly, these same cytokines had no impact on B-cell acute lymphoblastic leukemia cell survival in culture. Third, they demonstrated that resistant AML cells promote CAR T-cell exhaustion through ongoing antigen stimulation. Finally, and crucially, they found that exposure to myeloid-supporting cytokines did not directly impair CAR T-cell killing of AML blasts but rather endowed AML cells with relative CAR T-cell resistance. Single-cell RNA sequencing implicated cytokine-induced activation of pro-survival signaling pathways as the underlying mechanism.
Additional experiments confirmed JAK/STAT pathway activation, specifically STAT3 and STAT5, by myeloid-supporting cytokines in both culture and serial patient bone marrow samples. Phosphorylation of STAT5 leads to the activation of anti-apoptotic proteins. Indeed, incubation of AML blasts with GM-CSF not only increased STAT5 phosphorylation but also increased protein levels of BCL2.
In Brief
This beautiful work clearly elucidates the pathways by which CAR T-cell activation leads to cytokine secretion and clinical CRS, which in turn promote AML cell survival and contribute to clinical CAR T-cell failure. Although frustrating, this novel mechanism presents an opportunity: perhaps small molecular inhibitors of JAK/STAT and/or BCL2 could upend this signaling cascade and restore CAR T-cell functionality. The authors show that this is in fact possible, at least in vitro, using ruxolitinib and venetoclax. Certainly, clinical translation of these ideas is critical, and we can surely expect clinical trials to test these concepts in the near future. Intriguingly, the authors hypothesize that perhaps this same myeloid-supporting cytokine mechanism could be responsible for the heightened incidence of secondary myeloid malignancies following CD19 and B-cell maturation antigen-targeting CAR T-cell therapies. If proven, this could have therapeutic implications beyond CAR T-cell therapy for AML. In conclusion, we are one step closer to unlocking the potential of CAR T cells in AML, which, if realized, could benefit countless patients in need.
Disclosure Statement
Dr. Muffly indicated no relevant conflicts of interest.
